United States
Environmental Protection
Agency
Office of Air Quality
Planning and Standards
Research Triangle Park NC 27711
EPA-450/4-80-020
October 1980
Air
Guideline For Applying
The Airshed Model
To Urban Areas
-------
EPA-450/4-80-020
Guideline For Applying The
Airshed Model to Urban Areas
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Air, Noise, and Radiation
Office of Air Quality Planning and Standards
Monitoring and Data Analysis Division
Research Triangle Park, North Carolina 27711
October 1980
-------
This document 1s Issued by the Environmental Protection Agency to
report technical Information of Interest to a limited number of readers.
Copies are available as supplies permit from the Library Services Office
(MD-35), U.S. Environmental Protection Agency, Research Triangle Park,
NC 27711; or, for a nominal fee, from the National Technical Information
Service, 5285 Port Royal Road. Springfield, VA 22161.
Publication No. EPA-450/4-80-020
-------
ACKNOWLEDGEMENTS
This guideline was prepared by David E. Lay land of the Source Receptor
Analysis Branch of the Monitoring and Data Analysis Division of the Office
of A1r Quality Planning and Standards. John Summerhays, also of the Source
Receptor Analysis Branch, made Important contributions to the guideline.
E. L. Martinez and Dr. Henry S. Cole provided additional assistance.
Guidance and direction were given by Joseph A. Tikvart and David H. Barrett.
Steven D. Reynolds, Thomas W. Tesche, and Lawrence E. Reid of Systems Appli-
cations, Incorporated provided various materials which were adapted for
Inclusion 1n the guideline. Appreciation 1s extended to Zada Nelson for her
care and patience 1n preparing the manuscript.
-------
FOREWORD
High ozone levels are one of the most pervasive air pollution problems
facing our nation's cities. While the planning and analysis tools available
to grapple with the problem have been less than Ideal, the costs associated
with the types of control measures necessary to alleviate the problem are
high.
The need for credible modeling techniques for evaluating the relation-
ship between emission reductions and ozone levels on an urban scale 1s widely
recognized. A credible technique should possess at least two basic attributes,
First, 1t should account for all significant physical and chemical processes
and second, 1t should be capable of being verified.
Empirical techniques have been proposed and used for developing State
Implementation Plans called for by the Clean A1r Act. Unfortunately,
empirical techniques have not fully satisfied the criteria of technical
rigor and verlflability. Photochemical grid models on the other hand meet
both of these criteria. However, the benefits of using such models are
achieved at appreciable cost in terms of data requirements.
The physical and chemical processes which lead to high observed ozone
levels are complex. The relationship between precursor emissions and ozone
air quality 1s time and space dependent. Among the major factors affecting
the generation of ozone are:
r the quantity and the spatial and temporal distribution of emissions
of nitrogen oxides and volatile organlcs
*
- the chemical composition of the volatile organlcs emitted
- the chemical reactions among ozone, organlcs, and nitrogen oxides
11
-------
- the spatial and temporal nature of the wind field
- the stability of the atmosphere and the dynamics of the mixed layer
- the Intensity and diurnal variation of solar radiation
- the transport of ozone and Its precursors Into the urban area from
aloft and Immediately upwind
- the loss of ozone and precursors by surface uptake
The Environmental Sciences Research Laboratory of the Office of Research
and Development, U. S. Environmental Protection Agency, has supported a multi-
year research effort by Systems Applications, Inc. (SAI) of San Rafael,
California. This research has led to the development of the Airshed Model.
Initial model development efforts and applications to the Los Angeles area are
fully described 1n a 15-volume series of reports by Roth, et al., (1971) and
Reynolds, et al., (1973a). These reports are summarized 1n three papers by
Reynolds, et al., (1973b, 1974) and Roth, et al., (1974). Evaluation of the
model's predictions and Its components at that time showed that Improved treat-
ments of some physical and chemical processes were necessary. As a result,
another series of research efforts was carried out to Improve the model. These
efforts are described 1n a seven-volume series of reports by Jerskey and Seinfeld
(1976), Jerskey, et al., (1976), Klllus, et al., (1977), Lamb (1976), Lamb, et
al., (1977), L1u, et al., (1976a), and Reynolds, et al., (1976) and in papers by
111
-------
Lamb, Chen, and Seinfeld (1975), L1u and Seinfeld (1975), L1u, Whitney,
and Roth (1976), and Reynolds (1977).
Previous versions of the Airshed Model have been successfully applied 1n
Los Angeles, Denver, and Sacramento. These applications are reported 1n Reynolds,
^
et alJ, (1979), Reynolds, et al., (1978), and Anderson, et al., (1977). Future
applications are planned for Tulsa, Philadelphia, Washington, D. C., Baltimore,
New York City, and Boston.
Although this guideline 1s written in terms of applying the Airshed
Model, the same general principles apply to the application of any photo-
chemical grid model. Mention of the Airshed Model by name does not consti-
tute formal endorsement of or preference for this model at this time.
The purpose of this guideline 1s to familiarize air pollution control
agencies and others with the Airshed Model and its potential for use in State
Implementation Plan (SIP) development and for evaluating alternative strategies
for the control of photochemical oxidants. This is not a how-to document nor
a formal planning document. Rather, it 1s one which attempts to outline what
taskssare Involved 1n using the Airshed Model. With this information it 1s
hoped that control agencies contemplating a photochemical model application
will be able to effectively plan for such a project. Planning considera-
tions.will be unique for each individual application. For example, the extent
to which In-house versus contract resources are used must be considered on
a case by case basis. This document only.describes the kinds of tasks involved
and the overall level of resources required for an Airshed Model application.
1v
-------
TABLE OF CONTENTS
I. Principles of the Airshed Model .... 1
A. Model Concepts and Mechanics ..... 1
B. Treatment of Significant Processes . 8
1. Advection and Dispersion 8
2. Emissions 9
3. Chemistry 10
II. Overview of the Modeling Process 15
A. Role of Modeling 1n A1r Quality Planning . . 15
B. Selection of the Modeling Region 17
C. Collection of Aerometric and Emissions Data 18
D. Model Verification and Control Strategy Analysis 22
III. Data Needs 24
A. Source and Emission Inventories 24
1. Requirements for Source and Emissions Data 24
2. Emission Projections ' 36
3. Emissions Data Handling 40
B. A1r Quality Data 47
1. Requirements for Air Quality Data 47
2. Collection of A1r Quality Data 49
C. Meteorological Data 53
1. Requirements for Meteorological Data 54
2. Collection of Meteorological Data 56
IV. Preparation of Model Input Data 61
A. Day Selection Criteria 61
B. Data Input Files 62
1. Preparing the Meteorological Data 68
2. Preparing the Air Quality Data 70
3. Preparing the Emissions Data 74
V. Evaluation of Model Performance 75
A. Graphical Techniques 77
B. Statistical Techniques 81
C. Rever1f1 cation 84
VI. Model Application for Air Quality Planning 92
A. Preparation of Future Year Emissions 93
B. Future Year Model Simulation 94
1. Baseline Simulation 95
2. Control Strategy Simulation . 96
C. Interpretation of Model Results 106
VII. Resource Requirements for an Airshed Model Study Ill
A. Personnel Ill
B. Computer Facility 114
C. Project Timetable ..... 115
D. Overall Project Costs 118
References 124
Appendix. Technical Description of the Airshed Model A-l
-------
CHAPTER I. PRINCIPLES OF THE AIRSHED MODEL
A. MODEL CONCEPTS AND MECHANICS
•
The Airshed Model 1s a photochemical dispersion model designed to
simulate the concentrations of ozone and Its precursors (nitrogen oxides
and organlcs*) as they evolve during a day's episode. This Involves using
a detailed set of meteorological and emissions data 1n order to simulate the
emissions of precursors, atmospheric dispersion, and the chemical reactions
that generate ozone.
The basis of the mathematics for simulating concentrations of ozone
and precursors 1s the conservation of mass. The major processes which
cause changes 1n the amount of mass 1n a given air parcel are emissions,
advecflon and dispersion of mass Into and out of the parcel, and chemical
changes 1n the composition of the parcel . Thus photochemical dispersion
models are based on solving the following equation (restated 1n terms of
concentrations:
(mass 1n)-(mass out) due
Concentration « Initial concentration + /to advection and d1spers1on\
* parcel volume '
of emissions) + ch in C0ncentrat1on due
parcel volume to chemical transformation (1)
In the Airshed Model, the urban area 1s divided Into a three dimensional
grid system for which solutions of Equation (1) are derived Individ-
ually for each grid cell. This process 1s Illustrated 1n Figure 1-1.
The first term 1n Equation (1) 1s "concentration." The purpose
of running the Airshed Model 1s to estimate the concentration of ozone
The term "organlcs" Includes both simple hydrocarbons and oxygenated
organlcs. However, the term organlcs and the term hydrocarbons are used
Interchangeably.
1
-------
(a) SPECIFICATION OF THE GRID
Transport
Transport
B Transport
i
2 Transport
§-
r~
Transport
-* —
=
Chemistry
Elevated Emissions
A Transpo
Chemistry
Elevated Emissions
A Transpo
I
Chemistry
Elevated Emissions
* Transpo
I ,
Chemistry
Surface Emissions
I
rt
MMBBHM
•MBBM
rt
t
Transport
Transport
Transport
Transport
t
Top of Modeling Region
Top of Mixed Layer
Ground Surface
Surface Removal
1 to 10 kilometers
(b) ATMOSPHERIC PROCESSES TREATED IN A COLUMN OF GRID CELLS.
Figure 1-1. Schematic illustration of the grid used and atmospheric
processes treated in the airshed model (adapted from Reynolds,
Tesche, and Reid, 1978).
-------
In each grid cell for a particular day. This requires solving Equation (1)
not only for ozone but also for NO, N02 and five classes of organic com-
pounds. These five classes of organlcs roughly correspond to paraffins,
oleflns, aromatlcs, aldehydes and ethylene. The nature of the five classes
and the rationale for making the distinction 1s discussed later 1n this
chapter. Also discussed later are several other species Indirectly affect-
Ing ozone production which may optionally be simulated 1n the Airshed Model.
Thus a simulation using the Airshed Model Involves calculating the con-
centrations of each major species 1n each grid cell for each hour being
simulated. For example, a typical model run may simulate 14 hours (5 a.m.
to 7 p.m.) using a grid system of 1200 grid cells (15 cells east-west,
20 cells north-south, and 4 cells high) or more. Consequently, the Airshed
Model makes a large number of calculations in simulating the formation
of ozone on any given day.
The next term 1n Equation (1) 1s the Initial concentration. The
equation suggests that the concentration at the end of a time step equals
the concentration at the beginning of the time step plus the changes 1n con-
centration (due to advectlon, dispersion, emissions, and chemistry) that
occur during the time step. The length of a time step 1s usually six
minutes or less. The user must supply the Initial concentrations for the
first time step of the simulation. The concentrations calculated at the
end of the first time step are then used as the Initial concentrations
at the second time step. This process continues until the entire
desired time period has been simulated.
A typical time for starting simulations 1s 5 a.m. This 1s prior to
the morning rush hour and so concentrations are still fairly low. This
makes the model estimates less sensitive to Initial concentration measure-
ments and more sensitive to the chemical and physical processes being
3
-------
simulated during the day. Thus, the Airshed Model requires a concentration
at the beginning of the simulation, I.e., Initial conditions, for each species
(03, NO, N02, five classes of organics and any optional species) for each
grid cell.
The next term 1n Equation (1) represents the effects of winds and
turbulent mixing. From the perspective of an Individual grid cell, the
wind portion of this term 1s based on the difference between the pollutant
concentration transported 1n from upwind cells and the pollutant concentration
being transported out of the cell. The dispersion portion of this term
1s based on the concentration gradients between the cell and each of the
six neighboring cells (above, below, and four sides). From a broader per-
spective, 1t 1s also necessary to be able to estimate the mass which 1s
advected and diffused Into the modeling region as a whole. Thus, an Input
requirement of the Airshed Model 1s a set of concentrations at the boundary
of the modeling region, I.e., boundary conditions. More specifically, the
Airshed Model requires that estimates of the concentrations be Input for
all species (I.e., 0., NO, N0«, five classes of organics and any optional
species) for each hour simulated along the sides and at the top of the
modeling region.
It 1s generally desirable to define the modeling region such that the
upwind boundary 1s, 1n fact, upwind of the city, in order that the boundary
concentrations are relatively low. However, 1n many parts of the United
States, substantial concentrations of ozone and other pollutants are trans-
ported Into a city from long distances upwind. This pollutant transport 1s
reflected In the boundary concentrations used 1n the simulation. Thus, It 1s
often Important to obtain an accurate estimate of boundary concentrations.
4
-------
The third term on the right side of Equation (1) accounts for pollutant
emissions. This term 1s quite Important, since the purpose of using the
Airshed Model 1s to determine the relationship between emissions of hydro-
carbons and NO and the resulting concentrations of ozone. The model
n
requires the emissions of these species to be specified for each grid cell
for each hour that 1s simulated. Most of the emissions are added to the
lowest grid cell In the respective column of cells. However, the Air-
shed Model also has the capability of adding emissions from elevated point
sources to elevated grid cells. Thus, the Airshed Model Introduces the
emissions at the height at which they are actually found.
The final term 1n Equation (1) accounts for the effects of chemical
transformations. This term expresses the changes 1n chemical composition
of the atmosphere that occur as hydrocarbons and NOX react 1n the presence
of sunlight to form ozone. This 1s one of the most complex components of the
Airshed Model. Fortunately, the chemistry of the Airshed Model 1s fully
contained within the model, so the user 1s not required to supply any
Information about atmospheric chemistry.
The Airshed Model also has the option for simulating the concentration
of carbon monoxide. Since CO 1s relatively nonreactive, simulation of CO
concentrations 1s useful for checking the accuracy of the model's treatment
of advectlon and dispersion.
It should be clear from the above discussion that the Airshed Model
1s substantially different from the Gaussian dispersion models typically
used for Inert pollutants. First, the Airshed Model does not assume steady
state conditions, since ozone concentrations are a function of the complete
meteorological and chemical history of the air mass. Unlike most Inert
pollutant model applications, the Airshed Model considers the chemical
5
-------
Interactions of emissions from throughout the urban area. Also, the
Airshed Model does not use the empirically derived dispersion coefficients
(I.e., slgmas) Inherent 1n the Gaussian formulation. Instead, the Airshed
Model uses eddy d1ffus1v1ty coefficients to estimate dispersion.
The above discussion has outlined the concepts by which the major physical
and chemical processes are considered in the Airshed Model. In order to
discuss the mechanics of the Airshed Model, however, 1t 1s necessary to
discuss grlddlng. As mentioned previously, a typical application of the
Airshed Model might use a grid system of 15 cells by 20 cells horizontally by
4 cells vertically. Each cell has constant horizontal dimensions, say 4
kilometers by 4 kilometers. On the other hand, the vertical dimensions of the
cells vary 1n time according to changes in the mixing height. The user
determines how many cells will be used below the mixing height. In this
way the mixed layer 1s divided Into several cells of equal dimension. The
vertical dimensions of the cells typically range from 25 to 500 meters.
An Important option in the Airshed Model 1s to use cells above the mixed
layer. These cells become significant when the mixing height rises, thereby
entraining pollutants from these cells Into the mixed layer. This process,
known as fumigation, 1s a frequent occurrence 1n the late morning hours as
the mixing height rises rapidly. Early morning emissions from elevated point
sources become trapped in these cells and are then mixed to the ground later
in the day.
Figures I-2(a) through (c) show the various possibilities for speci-
fying the vertical dimensions of cells. Figure I-2(a) and I-2(b) show the
options of simulating the mixed layer with and without also simulating an
elevated Inversion layer. Figure I-2(c) shows that a third option of the
Airshed Model 1s to have the top of the modeling region below the mixing
height. The purpose of this option 1s to reduce the vertical dimensions of
6
-------
o
MIXING HEIGHT: REGION TOP
t /1 / 11 / 11 /
REGION TOP
MIXING HEIGHT
fftffftf iff n i i J
(a) MODELING REGION IS THE
SAME HEIGHT AS MIXED LAYER
(b) MODELING REGION INCLUDES
CELLS ABOVE MIXED LAYER
to
MIXING HEIGHT
r r^i^ T* t i r , •• *
REGION TOP
(c) MODELING REGION DOES NOT
INCLUDE THE ENTIRE MIXED LAYER
Figure 1-2. Several possible options for vertical gridding.
-------
the cells and thus Improve the vertical resolution. However, this option has
the disadvantage that the mixing height is not within the modeling region
making 1t difficult to properly simulate the effect of a I1d on the vertical
flow of pollutants.
B. TREATMENT OF SIGNIFICANT PROCESSES
The first part of this chapter provided a general discussion of the
concepts and mechanics of the Airshed Model. As stated previously, the
major processes considered 1n the model are advectlon and dispersion, emis-
sions, and chemistry. The following discussion 1s Intended to provide a
more detailed discussion of each of these processes.
1. Advectlon and Dispersion
Since the Airshed Model simulates concentrations 1n fixed grid cells,
advectlon and dispersion are simulated by transferring appropriate amounts
of pollutants from one cell to another. Pollutants are transferred horizontally
from one cell to the next as a function of the concentration and the speed of the
wind. The flux of material due to these factors 1s called advectlon. Pollutants
are transferred between two cells also as a function of the difference between
the concentrations 1n the two cells and an eddy d1ffus1v1ty coefficient. The
flux of material due to these factors 1s called dispersion. The eddy d1f-
fuslvlty coefficient expresses the rate of atmospheric mixing caused by
turbulent eddies.
The Airshed Model also simulates vertical advectlve transfer.
Vertical transfer arises primarily from two sources: (1) balancing the
effect of convergence and/or divergence of horizontal winds 1n order to
achieve a proper mass balance of air, and (2) accounting for changes 1n
cell height as the mixing height changes. The first source of vertical
8
-------
transfer must be considered 1n preparing the wind data for use 1n the
model. For example, 1f the horizontal winds are converging 1n a cell,
the wind field used 1n the model must have a sufficient vertical component
to prevent an artificial accumulation of pollutants 1n the cell. The
second source of vertical transfer 1s an artificial one arising from the
mechanics of the simulation program. As the cell height Increases, material
located at a given height may be transferred from one cell to another with-
out actually changing Its relative vertical position.
The Airshed Model has differences 1n Us treatments of vertical and
horizontal dispersion. Because horizontal dispersion 1s generally less
Important than advectlon, the Airshed Model uses a single horizontal eddy
d1ffus1v1ty coefficient. However, vertical eddy d1ffus1v1ty coefficients
are Individually calculated for each cell and each time period as a function
of atmospheric stability and surface roughness.
2. Emissions
The treatment of emissions 1s straightforward conceptually: the emis-
sions Into a cell are added uniformly throughout the cell. Thus, the change
1n concentration due to emissions simply equals the mass added divided
by the volume of the cell. Ground-level emissions are added Into the lowest
level cell. Elevated emissions are added Into upper level cells. This
requires that certain stack parameters be Input for the major point sources.
The stack parameters are used along with the wind speed to estimate an
effective plume height for each major point source. The emissions are
then added uniformly throughout the (upper level) cell which 1s calculated
to receive the plume.
-------
The Airshed Model 1s designed to treat emissions of eight pollutants:
NO, N02 five classes of organlcs and CO. NO emissions must therefore be
divided Into emissions of NO and N02 and organlcs must be divided Into emis-
sions of five organlcs classes. These classes are discussed 1n Section 3.
A point of comparison between the Airshed Model and Gaussian dis-
persion models 1s the treatment of plume rise. Like most Inert pollutant
dispersion models, the Airshed Model uses the Brlggs (1971) formulae to
estimate plume rise. However, unlike most Gaussian models, the Airshed
Model 1s capable of simulating emissions that rise above the mixed layer.
Inert pollutant models generally disregard any emissions 1n plumes that
rise above the mixing height. Although the Airshed Model also disregards
any emissions 1n plumes that rise above the top of the modeling region,
the modeling region 1n the Airshed Model can extend above the mixing
height. It 1s thus possible to simulate plumes that rise Into an Inversion
layer between the mixing height and the top of the modeling region.
3. Chemistry
As mentioned previously, the chemical mechanism 1s one of the most
complex components of the Airshed Model. The chemistry of ozone production
has been the subject of Intensive study for well over a decade, and yet
there 1s surprisingly little agreement on the 11st of reactions to use 1n
simulating ozone photochemistry. Thus, this discussion will only provide
a general outline of the photochemistry of ozone production. The discus-
sion will then outline the major distinguishing features of the chemical
mechanism used In the Airshed Model.
The general process by which tropospherk ozone 1s formed 1s Illus-
trated 1n Figure 1-3. Figure I-3(a) shows the N02-NO-03 cycle. These
reactions may be written as chemical equations:
10
-------
N02
SUNLIGHT
(a) THE N02-NO-03 CYCLE
N02
PHOTOCHEMICAL
BY-PRODUCTS
ORGANIC COMPOUNDS
FREE
RADICALS
(b) ORGANIC OXIDATION OF NO TO N02 WITH OZONE BUILDUP
Figure 1-3. Photochemical production of oxidants.
11
-------
a. N02 + sunlight -»• NO + 0
•
b. 0 + 02 -»• 03
c. 03 + NO -»• N02 + 02
Ozone 1s not emitted 1n any measurable quantity, and 1n fact, reactions
(a) and (b) represent the only significant source of tropospherlc ozone.
However, these reactions are fully reversed by reaction (c). Thus, NO by
n
Itself will not cause significant concentrations of ozone. The only way
significant concentrations of ozone will occur 1s for some other species
to oxidize NO to N02 without destroying ozone. This 1s exactly what 1s
shown 1n Figure I-3(b). This may be written as:
d. Organic radicals + NO -»• NOg + miscellaneous products
The net result of reactions (a), (b), and (d), then, 1s a recycling of
N02/N0, a modification of the organic species, and the generation of an
excess Og molecule. That 1s, organic species permit the ozone formation
step while bypassing the ozone destruction step of the N02 - NO - 03
cycle. The result 1s a buildup of the ozone concentration. Reactions
(a), (b), and (c) are fast enough that a balance or equilibrium 1s still
maintained between the concentrations of N09, NO, and 0*, but this balance
f Z 3
Involves a much higher ozone concentration than would occur without organlcs,
The above four reactions provide an overview of the chemical process
of ozone formation. However, the number of reactions which affect this
process 1s far larger. In fact, the number of reactions Involved 1s too
large for all the reactions to be Included 1n photochemical models. Thus
the chemical reactions simulated 1n these models are a simplified rep-
resentation of the Innumerable reactions that actually occur.
It 1s more difficult to represent the various reactions of organlcs
than 1t 1s to represent the reactions of NOX. There 1s substantial
12
-------
agreement on about 10 to 15 NO reactions as being the most significant
A
reactions for describing the fate of NO... However, there 1s relatively
^%
little agreement as to how best to represent organic reactions. This 1s
primarily due to the almost Infinite number of species and reactions that
occur. In tracing the entire path of decomposition of just one organic
species, 1t 1s possible to 11st over 100 reactions of that species and
the products of Its decomposition. Moreover, urban emissions Include a
wide variety of organic species. Therefore, a simplified representation
of all these reactions 1s needed.
The typical means of representing significant organic reactions 1s
to utilize a small number of reactions for about three to five categories
of organics. For example, the Uvermore Regional A1r Quality (LIRAQ)
model developed by the Lawrence Llvermore Laboratory at Llvermore, Cali-
fornia, (NacCracken and Sauter (1975), MacCracken et al., (1978)) uses
three categories of organics roughly described as oleflns, paraffins, and
aldehydes. Miscellaneous compounds are Included in the most similar of
the three categories. For each of these organics categories, LIRAQ uses
about 10 reactions. The reactions and the reaction rates are chosen to
represent the chemistry of an "average compound" in the respective chemical
category.
The means of representing organics reactions in the Airshed Model
1s somewhat different from that used 1n LIRAQ. In particular, the cate-
gories used In the Airshed Model represent total numbers of bonds of
specific bond types rather than total numbers of molecules of specific
molecular types. In other words, the Airshed Model treats Individual com-
pounds not as molecular units but rather as carbon-bond units. Several
different carbon-bond types are recognized: single bonds, double bonds,
13
-------
aromatic bonds, and carbonyl bonds. (These can be recognized as the
Identifying bond types for paraffins, oleflns, aromatlcs and aldehydes.)
The double bonds are further broken down Into highly reactive and moderately
reactive double bonds. The single bonds actually correspond to the number
of carbon atoms which are only singly bonded. An aromatic bond 1s actually
treated as an aromatic ring. Carbonyl bonds Include not only those
associated with aldehydes, but also those associated with ketones and esters.
The single bonds are the least reactive while the highly reactive double
bonds are, as their description Implies, the most reactive; the others are of
Intermediate reactivity.
Another species which can be considered 1n the Airshed Model Is CO.
Carbon monoxide plays only a minor role 1n ozone photochemistry and most
photochemical models consider it Insignificant enough to Ignore. Carbon
monoxide has more significance as a relatively Inert gas which may be used
as a tracer. By comparing the CO concentrations estimated by the model to
measurements of CO concentrations, it 1s possible to obtain a direct
assessment of the accuracy of the treatments of emissions, advectlon, and
dispersion without the complicating Influence of chemistry. This can be
quite useful for assessing model performance.
Another feature of the chemical mechanism 1s the option to consider
temperature effects. Temperature has a significant effect on the rates
at which many reactions occur, although 1t 1s unclear what effect 1t has on
overall ozone production. Unfortunately, there 1s little known about the
temperature dependence of the rates of many Important photochemical
reactions. Nevertheless, the Airshed Model does provide the option to
utilize current knowledge about the temperature variation of the rate
constants used 1n the chemical mechanism.
14
-------
CHAPTER II. OVERVIEW OF THE MODELING PROCESS
A. ROLE OF MODELING IN AIR QUALITY PLANNING
The Clean A1r Act requires the States to prepare State Implementation
Plans (SIPs) for assuring that ambient air quality standards are met within
a reasonable time frame by Implementing various control measures on a
progressive schedule. In the case of oxldants, the control measures thought
necessary Involve considerable expense to the community and may require
significant changes 1n transportation and land use practices. While there
are many elements Involved 1n achieving public acceptance of and support
for such measures, a fundamental element 1s the air quality planning and
analysis which lead to their adoption. Decision-makers, legislators, and
the concerned public must perceive the planning process as one 1n which
due consideration 1s given to the technical aspects of the oxldant problem.
Success hinges on credibility which In turn requires that recommendations
for action be based on a firm technical foundation.
Modeling 1s a tool which enables air quality planners to demonstrate
a cause and effect relationship between specific control measures and
anticipated Improvements 1n air quality. Photochemical grid models, such
as the Airshed Model, simulate the physical and chemical processes which
lead to observed oxldant levels in urban areas. Model results are
verified against measured ozone data. The effects of a given emission
reduction not only on peak ozone levels but also on Its spatial and
temporal distribution and on total population exposure can be determined.
Using such a model, the air quality planning agency will have the
15
-------
Information and tools at Its disposal to prepare and defend with confidence
Its plan for achieving the ozone standard and Improving public health.
Use of the Airshed Model 1s no small undertaking. The model requires
substantial Input data and lengthy computer runs to operate. A successful
modeling study depends on the active participation of the air pollution
control agency, the transportation planning agency, the regional planning
agency, and all affected city and county governments.
The experience of the Association of Bay Area Governments (metropolitan
San Francisco) 1n carrying out a detailed photochemical modeling study has
been summarized 1n a report entitled "Application of Photochemical Models 1n
the Development of State Implementation Plans, Volume I: The Use of Photo-
chemical Models 1n Urban 0x1dant Studies," (ABAG, 1979). Although a different
model, LIRAQ was used, the planning considerations, data collection activities,
and modeling aspects of the San Francisco application lend Insight Into the
use of photochemical dispersion models more generally.
While the actual model application and much of the data collection may be
done under contract, the study direction necessarily comes from the local
agencies. The local agencies are also a primary source of Information. The
development of a suitable emission Inventory for a base year along with pro-
jections of anticipated growth in future years requires the coordinated
efforts of the regional planning agency, the pollution control agency, and
the transportation planning agency. Because highway mobile sources are such
a major contributor to the oxldant problem 1n urban areas, the use by the
transportation agency of sophisticated transportation and emissions models
1s a critical element 1n the study. Collection of supplementary air quality
and meteorological data requires the Involvement of the pollution control
agency. The selection of control measures to be evaluated and the selection
16
-------
of a final strategy requires the participation of all these bodies of
government. Interaction of all affected groups will go far 1n assuring
that the modeling study meets the needs of the community and results 1n an
acceptable plan which, when Implemented, achieves air quality goals.
B. SELECTION OF THE MODELING REGION
Before a photochemical modeling study can begin, 1t 1s first necessary
to define the extent of the modeling region. In doing so, a number of criteria
should be observed. In simplest terms however, the modeling region must be big
enough to encompass the problem. Unfortunately, few urban areas have sufficient
j
ozone monitoring data to determine the spatial extent of elevated ozone levels.
The modeling region should encompass the urban core and outlying suburban
areas which together comprise the urban metropolitan area. It should also
Include significant satellite cities or towns and any major exurban Industrial
sources of oxldant precursors, e.g., hydrocarbon or nitrogen oxides. Allowance
should be made to Include areas where future growth 1s anticipated. In general,
the modeling region should be large enough so as to Include all major nearby
upwind sources under meteorological conditions known to be conducive to ele-
vated ozone levels.
Transport of oxldants and precursors Into the metropolitan area from
adjacent airsheds and formation of oxldants downwind of the metropolitan
area are also Important considerations when determining the extent of the
modeling region. Transported ozone can significantly affect ozone levels 1n
urban areas. As discussed 1n Chapter III, 1n order to quantify the Importance
of transport, 1t 1s necessary to locate supplementary ozone monitoring sites
well upwind of the city. Similarly, ozone monitoring sites downwind of the
urban area are necessary to determine ozone maxima resulting from ozone
17
-------
within the urban plume. In general, the modeling region should encompass
both the upwind and downwind monitoring sites.
While the modeling region must be large enough to adequately study the
urban oxidant problem, too large a region 1s undesirable. Emission Inven-
tory costs go up as the size of the modeling region Increases. In addition,
computer resource requirements for modeling are proportional to the total
number of grid squares 1n the modeling region. A compromise must be reached
between the size of the modeling region and the size of an Individual grid
square. Grid squares approximately 5 kilometers on a side are needed.
In general, the larger the grid size, the more localized effects are
smoothed over and the less useful Information 1s provided by the model.
By orienting the modeling region 1n an upwind-downwind direction, 1t may
be possible to reduce the size of the region while retaining the desired
grid size. Figure II-I shows the modeling region being used for an Air-
shed Model application 1n metropolitan Philadelphia.
C. COLLECTION OF AEROMETRIC AND EMISSIONS DATA
Use of the Airshed Model 1n an urban area requires a large aero-
»
metric and emissions data base. Existing data bases are almost universally
Inadequate 1n Important respects, thus necessitating the collection sup-
plementary emissions, air quality, and meteorological data. The required
spatial resolution of a modeling data base 1s on the order of 5 kilometer
grid squares while the required temporal resolution 1s on the order
of 1 hour. Data collection programs must be designed with these modeling
18
-------
1
/
q-
y
f
\
1*
—
I
]
r
—
^
— •
—
/
.
/
i
1
A
/
^,
f
/
^~'«
-^
J
^
'V
/
r
./••
/
\
i
/s
^(l
'"W
-
/*
^^
/
X
V
>
4
<
^.
c.
*
4
•;
.'
/
\
f
V^
*>
•\
X
/
i
V
T
r»
A
4
J
<
r
s
J\
^
^
V
r
t
s
N
\
"X.
t
x-
\,
/
{
f.s
•I
1
•*•%<
^
^>
T.
s
V1
\
\
^•*
s,
*>
V
,
S
\
L.
^~
^
i
I
•
/
/
i
^
\
^
I
t
>^
;
*~
\
^M
>v
N<
P
•
\
l
T
')
\
I
*^y
/
r
1*ii
tudj
(
^^
i
i
V-
\
7
iMn
ran
i
\
"X,
f
/
•^^
/
Ma
ta
.^
N
\
^"
\
f
Sti
•M
t.—
v9
A
»
^
rly
•»^
_^M
x
1
1
Aref
••• •
h
W
\
>
x»
\
^T
v
>
J
county Unas
Figure 11-1. Example of a Modeling Region and Grid Specification for an Airshed Model
Application (adapted from Engineering-Science, 1979b).
19
-------
needs 1n mind. The Airshed Model, as discussed 1n Chapter IV, offers extensive
data preparation programs which help make the best use of the data collected.
Collection of source and emissions data should focus on upgrading and
Improving the existing source/emission Inventories. While the Airshed Model
accepts all five criteria pollutants (carbon monoxide, sulfur dioxide, partlc-
ulates, nitrogen oxides, and hydrocarbons), emphasis should be placed on
sources of hydrocarbons and nitrogen oxides. As discussed 1n Chapter III,
three fundamental source types are recognized: point sources, area sources,
and line sources. Efforts should first be directed at developing a compre-
hensive annual Inventory of Individual point sources and distributed area sources
for the base year. Area sources must also be allocated to the grid network
selected for the modeling region. At the same time, a base-year, comprehensive
traffic Inventory Is needed for line sources. Vehicle miles associated with
both network traffic and off-network traffic must be accounted for.
Source by source Information on the species composition of hydrocarbon
emissions 1s a critical element of the emissions data collection effort. This
Information 1s used to allocate the total hydrocarbon emissions to the five
carbon-bond categories required by the Airshed Model. This allocation
effectively sets the reactive potential of the hydrocarbon emissions 1n
generating ozone.
Additional data collection 1s necessary to characterize the temporal
distribution of emissions. Both seasonal and diurnal variations should be
accounted for. Also, Information on anticipated growth on a regional
scale 1s needed to project the base year emissions to future years. The
final goal 1s an hourly emissions Inventory of all sources, for the base
20
-------
year and the projection years that 1s representative of a weekday during
the "oxldant season."
Collection of aerometrfc (meteorological and air quality) data should
»
focus on supplementing the existing monitoring network. This 1s generally
done In the form of a comprehensive "summer study." Such a study should
last about three months during the oxldant season so that a variety of
high-ozone episodes are represented. A1r quality and meteorological data
are gathered on a ground-based network consisting of both existing and
supplementary monitoring sites. Continuous ozone, nitrogen d1ox1de/n1tr1c
oxide, and nonmethane hydrocarbon monitors are used to characterize ground-
level air quality. Meteorological Instruments for measuring wind speed
and direction, temperature, and solar4 Insolation are variously collocated
with the air quality monitors. Rawlnsonde and plbal releases are used for
characterizing upper-level winds, mixing height, and stability. In addition,
aircraft overflights are desirable 1n some metropolitan areas in order to
characterize air quality aloft. Finally, grab samples are collected and
analyzed for Individual hydrocarbon species in order to further characterize
the ambient reactive mix.
As discussed 1n Chapter III, the air quality data 1s needed both to
specify the Initial and boundary conditions for model application and to
evaluate and verify the model's performance at estimating ozone levels in
the base year. The meteorological data 1s necessary to prepare mixing
heights and a three-dimensional wind field which, together with emissions,
are the major driving forces which control the model predictions.
21
-------
D. MODEL VERIFICATION AND CONTROL STRATEGY ANALYSIS
One of the key features of using a photochemical grid model such as the
Airshed Model 1s that model results are verifiable: model estimates of ambient
ozone concentrations can be compared directly with observed concentrations.
Various statistical and graphical performance measures may then be employed
to assess whether the model Is performing satisfactorily.
Model verification 1s conducted for the base year. In other words, the
base year emission Inventory together with the aerometrlc data collected dur-
ing the base year summer study are used to prepare the model Input data. A
total of eleven data files must be prepared prior to running the simulation
program; several other nondata files are also required. The Airshed Model
Includes a variety of data preparation programs. However, not all the methods
and options offered may be sufficient or appropriate for a particular appHca-
w
tlon, necessitating modifications of the existing programs or writing of new ones,
Several days during the summer study period should be selected for model
verification. Such factors as ozone levels and data completeness should be
considered, among others. Simulation of more than a single day allows one
to evaluate how the model performs for a range of conditions under which
high ozone concentrations occur ! Moreover, satisfactory model performance
•I
may not be achievable on a particular day causing one to turn to other days.
The emissions data base 1s not normally developed for a specific calendar
day, but rather for a representative weekday. However, the meteorological
and air quality data must be prepared separately for each calendar day for
which a simulation 1s to be run.
Model verification requires a high level of skill and judgment on the
part of the modeler In addition to an Intimate familiarity with the data base.
22
-------
Verification 1s an Iterative process whereby model executions are
followed by an evaluation of the model's performance, a diagnosis of
any problems, and a reanalysls of the model Input data. Modifications
are made, the model 1s rerun, and the process 1s continued until satis-
factory performance 1s achieved on one or more days.
Once the model and data base have been verified, the Airshed Model
1s ready to be applied to the analysis of control strategy options. This
step Is the culmination of all previous steps where the benefits of the
photochemical modeling study are realized. The question of what control
measures should be Imposed on which sources with what level of stringency
may be answered using the Airshed Model as an analysis tool.
Control strategies are evaluated for a critical test day when the
mesoscale meteorology, pollutant transport, and pollutant carryover (from
previous days) result 1n high ozone levels. Control strategies may be
formulated using a sensitivity approach to determine an overall regional
emission reduction or by a more rigorous approach 1n which candidate stra-
tegies are simulated. Control strategy evaluation should be conducted for
several test days of different meteorological and air quality data. The
effectiveness of a particular control strategy may vary depending on the
aerometrlc conditions on a particular test day. The ultimate goal 1s a
control strategy consisting of specific control measures for specific
sources which, by means of model simulation, 1s shown to attain the ozone
standard on each of the high-ozone test days.
23
-------
CHAPTER III. DATA NEEDS
A. SOURCE AND EMISSION INVENTORIES
The Airshed Model and other photochemical grid models attempt to
simulate the production, decay, and transformation of multiple pollutants
1n a reactive atmosphere. In order to portray this reactive mix over time,
1t 1s necessary to estimate the total emissions of photochemlcally reactive
pollutants and their spatial and temporal distribution. The principal
reactive pollutants affecting ozone formation are hydrocarbons and nitrogen
oxides. The Airshed Model requires an hourly breakdown of the total
ground-based emissions and elevated emissions of these pollutants for
each grid square. In addition, emissions of carbon monoxide are also
useful during the modeling study.
1. Requirements for Source and Emissions Data
As Indicated In Table III-l, emission sources are generally broken down
Into three categories: line sources, area sources, and point sources. Line
sources consist of motor vehicle emissions from streets and highways. In terms
of transportation planning, line sources encompass both link VMT (vehicle
miles traveled) associated with vehicle trips on the transportation network
and intrazonal VMT associated with trips off the network. Line sources are the
single most Important source of precursor emissions in urban areas. Area
sources consist of numerous small stationary source emissions and off-highway
mobile source emissions. Point sources consist of stationary sources of vary-
ing size which are Individually Identified in detail. While the Airshed
Model Itself requires only emission rates, collection and storage of the source
24
-------
Table III-1
Source and Emissions Data Needs
Category
Line Sources
Data
Network Links
hourly link emissions;* endpolnt
locations; vehicle mix and model year
mix; diurnal pattern of VMT, average
speed, and % hot and cold starts; other
auxiliary data; VOC and NOX species data,
Intrazonals (off-network)
hourly zone emissions;* grid
allocation factors; other data as above.
Area Sources
Point Sources
annual county-level emissions* and
activity levels; grid allocation factors;
temporal apportioning factors, VOC
and NOX species data.
plant Identification; point Identification;
annual point emissions;* locations; stack
parameters; control efficiencies; process
Identification and operating rates; temporal
apportioning factors; VOC and NO species
data. x
Emissions of HC, NOX, and CO.
25
-------
data Indicated 1n Table III-l 1s needed to perform the computations necessary
to project emissions to future years, both to examine the effect of growth
and of additional control measures, and to prepare the data for model Input.
The necessary procedures for developing the kinds of data shown 1n Table III-l
have been discussed elsewhere. "Procedures for the Preparation of Emission
Inventories for Volatile Organic Compounds, Volume I" (EPA, 1977b) describes
the methods used for compiling annual Inventories of VOC emissions from point and
area sources. However, the methods for Inventorying area sources are Intended
for relatively large geographic areas, such as counties. A companion document,
Volume II, 1s entitled "Emission Inventory Requirements for Photochemical A1r
Quality Simulation Models," (Pacific Environmental Services, 1979). It describes
requirements and methodologies for developing an Inventory of point, area, and
line (I.e., highway mobile) sources for use 1n photochemical grid models, such
as the Airshed Model. For point and area sources, the spatial and temporal
resolution in Volume II builds on the basic annual Inventory of Volume I. How-
ever,-for highway mobile sources, a completely different procedure 1s recom-
mended which 1s based on the use of transportation models and network emission
models. Volume II also provides a general description of VOC species data. How-
ever, the use of such data for an Airshed Model application 1s described here.
The following provides an overview of the Inventory process and highlights the
salient features an Inventory should possess 1f it 1s to be used for an Airshed
Model application.
In general, a metropolitan area planning to undertake a photochemical model-
1ng study should already have most of the basic source/emissions Inventory
Information 1n some form. However, 1t 1s usually necessary to update the point
and area source Inventories, assemble the various data elements for the line
26
-------
source Inventory, develop grid allocation factors for area sources, and conduct
original surveys of point or area source categories not currently accounted for.
A good annual Inventory 1s the necessary starting point for area and point
sources while a good transportation Inventory 1s necessary for Hne sources.
Updating of the point and area source Inventories should focus on sources
of hydrocarbons and nitrogen oxides. Table II1-2 lists the principal sources
of these pollutants 1n urban areas. While the point source Inventory
Identifies Individual, discrete processes, the area source category break-
down used depends upon the relative similarity of different sources. Dif-
ferent allocation parameters, temporal patterns, organic species compositions,
or emission factors may dictate subdivision of the categories listed 1n
Table II1-2. For example, the different solvents used for drycleaning have
different chemical compositions and different reactivities. When allocating
the solvent emissions to the five carbon-bond categories, a stoddard solvent
1s treated differently than perchoroethylene. In addition, consideration
should be given to the control measures to be evaluated. For example,
1f stage I and stage II service station controls are to be analyzed
separately, then it may be desirable to subdivide service station emissions
Into those associated with loading underground tanks and those associated
with automobile tank filling.
Once the point, area, and line source Inventories are assembled, the
emissions must be apportioned to hourly periods. The hourly emissions
should be representative of a typical weekday during the oxldant season
(usually June through September). Original surveys of hourly process rates
for the major point sources of hydrocarbons and nitrogen oxides are desirable.
For smaller point sources, plant operating schedules are sufficient to derive
typical hourly emissions, while for area sources temporal patterns for each
27
-------
Table III-2
Sources of Hydrocarbon and Nitrogen Oxide Emissions
PETROLEUM REFINERIES
Process Drains and Wastewater Separators
Vacuum Producing Systems
Process Unit Turnaround
Boilers and Process Heaters
Catalytic Cracking
Chemical Treating
Coking
Blending
A1r Blowing
Fugitive Sources
STORAGE, TRANSPORTATION, AND MARKETING OF PETROLEUM PRODUCTS
011 and Gas Production Fields
Natural Gas and Natural Gasoline Processing Plants
Pipelines
Petroleum Product Storage
Ship and Barge Transfer.of Gasoline and Crude 011
Bulk Gasoline Terminals
Gasoline Bulk Plants d
Service Station Loading
Service Station Unloading
ORGANIC CHEMICAL MANUFACTUREf
Process Streams
Storage and Handling
Waste Handling
Fugitive Sources
INDUSTRIAL PROCESSES
Paint, Varnish, and Ink Manufacture
Food Processing
Pharmaceutical Manufacture
Rubber Products Manufacture
Plastic Products Manufacture
Textile Polymers Manufacture
Coke Production
Mineral Products Manufacture
Nitrate Fertilizer Manufacture
Nitric Add Manufteture
Explosives Manufacture
28
-------
Table III-2 (continued)
INDUSTRIAL SURFACE COATING
Large Appliances
Magnet Wire
Automobiles
Cans
Metal Colls
Paper
Fabrics
Metal Furniture
Wood Furniture
Flat Wood Products
Other Metal Products
NONINDUSTRIAL SURFACE COATING
Architectural Coatings
Auto Reflnlshlng
OTHER SOLVENT USE
Degreesing (toluene, trlchloroethylene, and other)
Dry Cleaning (stoddard, perchloroethylene, and other)
Graphic Arts
Adheslves
Cutback Asphalt Paving
Miscellaneous Commercial/Consumer Solvent
Pesticides
STATIONARY SOURCE COMBUSTION
External Fuel Combustion
Electric Generation
Industrial Boilers
Commercial and Institutional Boilers
Residential Heating
Internal Fuel Combustion
Turbine
Reciprocating Engine
Solid Waste Combustion
On-Site Incineration
Open Burning
HIGHWAY MOBILE SOURCES
Link VMT
Light Duty Autos
Light Duty Trucks
Heavy Duty Gasoline Trucks
Heavy Duty Diesel Trucks
Motorcycles
Intrazonal VMT
Same as above
29
-------
Table II1-2 (continued)
OFF-HIGHWAY MOBILE SOURCES
Railroads
Aircraft
Vessels
Construction Equipment
Agricultural Equipment
Other Vehicles
aPetroleum Product Storage—Includes all storage facilities
except those at service stations and gasoline bulk plants.
Bulk Terminals—emissions from loading tank trucks and
rail cars.
°Bulk Plants—emissions from storage and transfer.
Service Station Loading—filling of underground storage tanks.
eServ1ce Station Unloading—automobile tank filling.
Organic Chemical Manufacture—only generic classes of processes
associated with the manufacture of various organic chemicals are listed.
30
-------
type of activity are needed. For Hne sources, hour-by-hour traffic counts
representative of various roadway functional classes (freeway versus arterial,
etc.) and locale (central business district versus suburban residential, etc.)
are needed to characterize the diurnal distribution of VMT.
Except when stack tests or materials balance calculations show otherwise,
standard AP-42 (EPA, 1977a) emission factors can usually be used for esti-
mating emissions for point and area sources. Occasionally other references
*
may be required for certain source categories. For line sources, the MOBILE1
computer program (EPA, 1978a&b) should be used for generating mobile source
emission factors. Most point and area source emission factors are assumed
constant with time. However, emission factors for highway mobile sources
vary dlurnally as vehicle operating conditions change over time. Both
average vehicle speed (as correlated with traffic loading) and percent hot
and cold starts show distinct variations from hour to hour and have a marked
effect on motor vehicle emissions. Therefore, hourly variations in emissions
from links and intrazonals are a result of hourly changes 1n both VMT and
emission factors.
The Airshed Model requires that hydrocarbon emissions be apportioned to
the five carbon-bond categories discussed in Chapter I in order to properly
treat the varying reactivity of different organic species. It is therefore
necessary to obtain organic species (VOC) data for each source or source
type. Publications are available which give typical organic profiles for a
wide variety of point and area source types in terms of the weight or volume
percent of the Individual organic compounds present [Bucon, Macko, and Taback
(1978); GHscom (1978); THjonls and Arledge (1976)]. Highway mobile source
exhaust profiles depend on whether the vehicle is catalyst equipped and there-
fore on the type of fuel burned. Exhaust profiles are obtained from the
*A revision of MOBILE1 1s expected in July, 1980.
31
-------
same publications as are profiles for the evaporative portion of gasoline
vehicle emissions and dlesel vehicle exhaust emissions.
Despite the utility of these publications, probably the best data, at
least for point sources, would be obtained from responses to questionnaires
sent out to individual sources. Much of the published data 1s for the Los
Angeles area and may not be applicable to other parts of the United States.
For example, surface coating operations exhibit large variations depending
on the solvent or coating used. In addition, gasoline composition 1s sig-
nificantly different in California than it is elsewhere. Profiles for gasoline
storage and handling, as well as for highway vehicles, should reflect the
local average gasoline composition. In the case of highway vehicles, published
profiles may be adjusted depending on the composition of the gasoline (Engineering-
Science, 1979). For sources where no data are available at all, an alternative
may be to use the urban ambient mix of organic species (as discussed in Section
B. A1r Quality Data of this chapter) as a surrogate profile.
A note of caution regarding hydrocarbon emissions 1s in order. Most
hydrocarbon emission factors reported 1n the standard reference AP-42 do not
account for aldehyde emissions. A few source types have aldehyde emission
factors reported separately 1n AP-42, but most do not. Therefore, 1t 1s
generally necessary to adjust the emissions of sources which emit aldehydes,
primarily stationary and mobile source combustion, to obtain total hydro-
carbon emissions. This may be done as follows:
Adjusted HC Emissions = HC Emissions ( — *)
I00-we1ght % aldehyde
*as formaldehyde (HCHO)
This adjustment 1s made prior to apportioning the emissions to the five
carbon bond categories, which 1s discussed below.
32
-------
The data needed to develop a profile 1n terms of the five carbon bond
categories used by the Airshed Model 1s (1) the weight percent and (2) the
molecular weight of each species present. If the data are reported 1n terms
of volume percents, 1t may be readily converted to weight percents. The
data are then used to compute first, the number of moles of each species
present, and second, the number of moles of each bond type associated with
each species. Once the moles of bond types are computed for all species,
they are summed to obtain the total number of moles of each bond type. The
data manipulations Involved are described further 1n the Appendix, Technical
Description of the Airshed Model.
While the technique just described Is recommended when detailed species
by species data 1s available, 1t may be found that some summarization of the
data by organic class has been undertaken prior to reporting the data. If so,
the calculations necessary to obtain the five carbon bond categories will
depend on the classes actually reported. For example, 1f the data have been
summarized 1n terms of paraffins, oleflns, aldehydes, ketones, alcohols,
acetates and aromatics, the equations in Table II1-3 could be used to approxi-
mate the carbon-bond categories.
The Airshed Model also requires that nitrogen oxides be apportioned
to the two species, NO and NOg. Little data are available regarding the
relative proportion of nitric oxide and nitrogen dioxide 1n emissions from
combustion sources. However, some data for stationary sources has been
published (MHHgan, et al., 1979) and data used 1n preparing a photo-
chemical Inventory for Tulsa have been reported (Engineering-Science, 1979).
While the above data Items are needed for each source to derive a
temporally resolved Inventory of the various pollutants required by the
Airshed Model, further data 1s needed to provide the appropriate level of
33
-------
Table III-3. Possible Equations for Estimating Airshed
Model Carbon Bond Categories
Carbon Bond Category Equation
mass oleflns C3+
MW oleflns C3+
(1) OLE =
i9\ ADO - mass aromatlcs
(Z) ARO ~ MW aromatlcs
t^\ rABR - mass aldehydes . mass ketones . mass acetates
\si WWD - MW aldehydesMW ketonesMW acetates
fA\ Dflo _ mass paraffins /MW paraffins - 2\
l4' PAR " fin
-------
spatial resolution. As discussed previously, the Airshed Model requires,
1n general, a grldded Inventory. However, the model will accept Individual
point sources. It 1s not necessary or desirable to treat all point sources
Individually for modeling purposes. Rather, point sources should be
separated Into "major" sources and "minor" sources. Only the major point
sources are left as Individual sources. The minor point sources are assigned
to grids and their emissions are aggregated. Two criteria should be observed
when selecting the major point sources. First, the emissions should be
released Into an elevated layer, at least 50 meters or more above the surface.
Second, the quantity of hydrocarbon or nitrogen oxide emissions should rep-
resent a significant fraction of the regional total emissions, on the order
of one half of one percent. In this regard, 1t 1s not so much the emissions
from an Individual source, but rather the total amount of elevated emissions
within a column of grid cells that 1s Important.
As with minor point sources, line sources and area sources must be
assigned to the grid squares. Link emissions are readily assigned on the
basis of their endpolnt coordinates. Where a link falls in two or more grids,
the emissions may be apportioned on the basis of the fraction of the length
of the link In each grid. Intrazonal line source emissions may be apportioned
according to the fraction of the area of the zone in each grid.
A variety of surrogate Indicators are used to apportion county area
source emissions to the grid squares. For example, population might be
used for apportioning domestic solvent emissions, commercial land use for
gasoline handling, and basic employment (as compared with service employ-
ment) for degreasing emissions. In this example, population, commercial
land use, and basic employment are termed allocation parameters. The
fraction of the county total of each parameter associated with each grid
35
-------
square 1s an allocation factor. Obviously, this approach presumes that data
for the allocation parameters are available at a "sub-county" level. Census
data along with transportation, land use, and economic studies oftentimes
provide such data for Individual analysis zones of sufficient spatial resolu-
tion for use 1n photochemical modeling. Of course, 1f the specific locations
of area sources are known, the use of surrogates 1s unnecessary. This may
*
be the case for dlesel locomotives, where track mileage and switchyard
locations are known, aircraft emissions where airport locations and relative
levels of activity are known, and gasoline handling where a survey of service
stations has been undertaken.
2. Emission Projections
A good base year emission Inventory as described above is necessary
in order to verify the model's performance (Chapter V, Evaluation of
Model Performance). However, Its primary importance is actually as a
starting point from which to project future emissions. Emission projections
serve as the basis for systematic planning to attain air quality goals. Two
types of projected inventories are recognized: (1) baseline projection
inventories, and (2) strategy projection Inventories.
Baseline projections Include the effect of growth and control regula-
tions now 1n effect. Their purpose 1s to estimate where an urban area will
be 1n terms of emissions and air quality at some future date 1f no additional
controls are adopted. The need for additional controls can then be assessed.
While no one has a crystal ball, 1t 1s possible with suitable techniques
and assumptions to estimate future emissions 1n a meaningful way.
36
-------
Baseline projections are usually done separately for point, area, and line
sources. Point and area source projections are generally the responsibility of
the air pollution control agency while line source projections should be done
by the transportation planning agency. Coordination 1s necessary to assure
that projections are done on a consistent basis.
For consistency, all growth projections, whether point, area, or line,
should be derived from the same basic social and economic Indicators. Such
Indicators Include population, employment, housing and land use. Regional
growth models, such as IPEF (Interactive Population/Employment Forecasting
Model) are used to forecast regional population and employment. These
forecasts are then used together with Information on land use, transportation,
and planning policy to forecast land use, population, employment, and housing
1n Individual transportation zones with a model such as PLUM (Projectlve
Land Use Model). These models are available from the Federal Highway Admin-
istration and are often used by transportation planners as basic Inputs to
the transportation modeling process. For projections of stationary source
emissions such forecasts are Invaluable, not only 1n and of themselves, but
also as regional control totals. Ideally, projections of both mobile and
stationary source emissions, at a regional level, should be firmly tied to
these fundamental Indicators of growth.
Transportation models use these social and economic Indicators to
estimate trip generation and distribution 1n future years. Once these
trips are assigned to a planned future highway network, the resulting
traffic loadings can be used to calculate emissions. (If the transportation
37
-------
model has not been run for precisely the same year(s) which Interest air
quality planners, 1t may be possible to Interpolate from the results for
years which have been run. If this 1s attempted however, 1t should be done
by the transportation planning agency to assure Its validity.) Revised
vehicle speeds appropriate for loadings on the future network should be
used 1n selecting emission factors, as should revised percent cold starts
and hot starts in areas where land use 1s projected to shift substantially.
Point and area source projections are usually done on a source category
by source category basis using the best Information available. Changes 1n
gasoline marketing might be estimated on the basis of total projected VMT
(from the transportation model) and average projected fuel economy. Changes
in petroleum refining might be based on a judgmental application of national
trends, on projected changes 1n local basic employment, or on the basis of a
local survey of the Industry. Unless Information to the contrary 1s available,
1t is generally necessary to assume that all emission points 1n a large
Industrial facility "grow" by the same factor. Only for a few sources, such
as power plants, 1s sufficient Information generally available on planned new
sources or facility expansions to make possible the actual addition of new
emission points to the Inventory.
Normally overall growth and the effect of control measures on area source
emissions are projected at the county level (or at a redefined "political
jurisdiction" level) while changes in area source spatial patterns are handled
through the use of allocation factors developed specifically for a certain
projection year. Most area source categories are projected using such
surrogates as population, employment and land use forecasts and are then
38
-------
allocated to grids using the same surrogates. It 1s therefore possible to
develop corresponding allocation factors for a specific projection year.
Strategy projections, as discussed 1n Chapter VI, Model Application
for A1r Quality Planning, are used to evaluate the effectiveness of addi-
tional proposed control measures and to test the final strategy selected
for attainment of the ozone standard. Typical control measures that may be
considered Include reasonably available control technology (RACT), lowest
achievable emission rate (LAEff), and best available retrofit technology (BART)
for point and area sources. Inspection and maintenance (I/M) and transporta-
tion control measures (TCMs) may be considered for line sources.
The level of control represented by RACT, for example, when applied to
a given source category, may be expressed as a control factor which 1s then
used to reduce the projected (baseline) emissions to arrive at the controlled
(strategy) emissions. When controls already exist, 1t 1s probably best to
start from the uncontrolled level and then apply the appropriate reduction.
A complication arises when, for example, 1t 1s desired to apply a level of
control corresponding to LAER to new sources while applying a level cor-
responding to RACT to existing sources of the same category. If growth 1s
anticipated, yet specific Information on new units does not exist or 1s
unavailable, 1t may be necessary to make some simplifying assumptions about
how much of the growth 1s "new" and how much 1s due to an Increase within
existing capacity and therefore "old." When making longer-term projections,
say to 1995, replacement of existing capacity may be another complicating factor,
While the simplifications and assumptions required are not Insignifi-
cant, the degree of detail with which projections are performed should be
commensurate with the accuracy of the growth Indicators from which they are
derived. Moreover, what 1s Important from a modeling viewpoint 1s the total
regional change 1n emissions and Its spatial and temporal distribution.
39
-------
3. Emissions Data Handling
The above discussion has outlined the wide variety of data required for
the base year and projection year Inventories. This data 1s used to generate
an hourly Inventory of eight pollutants (NO, N02, five carbon-bond categories,
and CO) for each grid square 1n the modeling region along with an Inventory
of hourly emissions for each major point source.
Assume for the moment that the modeling region contains 400 grid squares
and 50 major point sources. For a fourteen hour simulation the Airshed Model
requires emissions for eight pollutants for each hour. The number of emis-
sion values Input to the model 1s therefore 400 grids x 8 pollutants x 14
hours/day plus 50 points x 8 pollutants x 14 hours/day or a total of 50,400
values. Each of these 50,400 values 1s generated from a series of calcula-
tions. These 50,400 values would then have to be regenerated for each base-
line projection and strategy projection to be simulated. Obviously, some
form of computer assisted procedure 1s necessary to perform the literally
millions of calculations and manipulations necessary.
A schematic representation of a computerized procedure for performing
the necessary calculations and converting the results to model format 1s
shown 1n Figure III-l. The procedure requires as Input an EIS/P&R file of
point and area source emissions. Due to Its design, the EIS/P&R system
(EPA, 1975) 1s quite flexible and well suited as the starting point from
which an Airshed Inventory of point and area source emissions may be
developed. Line sources should be handled separately however; this 1s
discussed later.
40
-------
REPORTS
REPORTS
TEMPORAL
AND
POLLUTANT
SPLIT
TEMPORAL
FACTORS
POLLUTANT
PROFILES
EIS/MR
HOURLY SOURCE/
EMISSIONS FILE
MAJOR/MINOR
POINT SOURCE
DIVISION/MODEL
CONVERSION
CONTROL
CONTROL
AREASOURCE
QRIDDINOmOOEL
CONVERSION
AREA SOURCES
TIME INTERVAL/
GRID VALUES
REPORTS I
Figure III-1. A POINT AND AREA SOURCE DATA HANDLING PROCEDURE.
41
-------
The EIS/P&R file has a hierarchical structure. Each facility or
"plant" may contain several "points" defined as stacks or boilers. Each
point may contain several processes or "SCCs." The SCC or Source Clas-
sification Code (EPA, 1976) Identifies the kind of process venting to a
stack or, 1n the case of boilers, the type of boiler and the fuel burned.
For point sources, stack data, operating schedule Information, control
efficiency, and annual estimated emissions are stored at the point level
while the annual fuel/process operating rate and emission factor are stored
at the process level. Area sources are readily treated as "pseudo-point
sources." In other words, the estimated annual county emissions from each
area source category are stored on the file at the point level. Such an
EIS/P&R "master file" 1s required for the base year and for each baseline
projection or strategy projection.
The manipulations necessary to generate the Airshed inventory are the
same regardless of whether the EIS/P&R master file represents the base year
file or one of several projection files. The procedure Itself consists of
three programs each of which performs several major functions. The first
program shown 1n Figure III-l performs two functions: (1) the annual
i
emissions are apportioned to 24 consecutive one-hour periods and (2) the
hydrocarbon emissions are apportioned to the five carbon-bond categories and
the nitrogen oxide emissions are split Into NO and NOg. Two Input files are
required to perform these functions: a temporal factor file and a pollutant
profile file. The temporal factor file contains a seasonal apportioning
factor, a daily apportioning factor, and a set of 24 hourly apportioning
factors for each source. In the case of a point source for which no temporal
42
-------
factors are Input, the operating schedule Information on the EIS/P&R master
file 1s used Instead. The pollutant profiles may be Input source by source,
or by source category. The profiles contain the weight percent of NO and N02
and weight percents for each of the five carbon-bond categories. (The develop-
ment of these percents 1s described 1n the Appendix, Technical Description of
the Airshed Model.) The pollutant profile file also contains the weight per-
cent of aldehydes. This 1s used to perform the aldehyde correction discussed
1n Section 1. Requirements for Source and Emissions Data.
The second program 1n Figure III-l divides up the pdlnt sources Into
major sources and minor sources. The user specifies a ton per year cut-off
for both NO and HC and a plume height cut-off. Point sources exceeding
A
these cut-offs then become major sources. The program then takes the major
sources and creates two files in Airshed Model format, one containing the
stack data for each source and the other containing the corresponding hour-
by-hour emissions. The minor sources are assigned to the modeling region
grid, the grid being defined by the user, and a file 1s created in model
format containing hour-by-hour emissions for each grid square.
The third program takes the county area source emissions and dis-
aggregates it Into gridded emissions. The user Inputs a file containing
a set of allocation factors for each allocation parameter. These factors
represent the fraction of each parameter, e.g., population, in each grid
square. The user also specifies which allocation parameter 1s to be used
for allocating which area source category, e.g., population, for domestic
solvent use. The program then creates a file in model format containing
hour-by-hour area source emissions for each grid square.
43
-------
While a data handling procedure of the type just described 1s suitable
for point and area source emissions, a different procedure 1s required for
line sources. Figure III-2 shows a conceptual flow diagram for such a
procedure. The procedure consists of several parts. First, as shown
across the top of the figure, link emissions are calculated. Intrazonal
emissions are also calculated and then combined with the link emissions when
the emissions are allocated to grid squares.
The computation of link emissions begins with a file of transportation
network data, a Federal Highway Administration (FHWA) "historical record"
file for example. Diurnal traffic distributions are then used to compute
hourly link traffic volumes. These volumes are used along with link capacity
to estimate hour-by-hour link speeds. The volumes are also used to compute
hour-by-hour VMT. This data, along with the other transportation data, 1s
stored on the "Hourly Link File." Next, the MOBI LEI program 1s used to compute
hour-by-hour emission factors for each link. Speeds are obtained from the
"Hourly Link File" while percent cold starts and hot starts, ambient tempera-
ture, and a variety of other Input parameters required by MOBILE1 are Input
separately. The percent cold starts and hot starts are Input for each
traffic zone, land use classification, or locale type. A preprocessor
program selects the necessary MOBILE! parameters for each link. The result
1s an hour-by-hour, I1nk-by-l1nk emission factor file. By multiplying these
emission factors times the hour-by-hour VMTs from the "Hourly Link File,"
the hourly link emissions are computed. A similar procedure 1s used to
compute Intrazonal emissions.
Because evaporative emissions have a distinctly different organic
composition than do exhaust emissions, evaporative and exhaust emissions
44
-------
CJ1
BASE OR
PROJECTION
YEAR
LINK
FILE
CALCULATE
HOURLY
LINK DATA
CALCULATE
HOURLY LINK
EMISSIONS
HOURLY
LINK
FILE
HOURLY
TRAFFIC
DISTRIBUTION
HOURLY
LINK
EMISSION
FACTOR
FILE
CONTROL
STRATEGY
OTHER
TRAFFIC DATA
BASE OR
PROJECTION
YEAR
INTRAZONAL
FILE
CALCULATE
HOURLY ZONE
EMISSIONS
GRID LINK
AND ZONE
EMISSIONS
HOURLY
ZONE
EMISSIONS
GRID
COORDINATES AND
ALLOCATION
FACTORS
CONTROL
STRATEGY
HOURLY
ZONE
EMISSION
FACTOR
POLLUTANT
PROFILES
OTHER
TRAFFIC DATA
( LINE SOURCES
TIME INTERVAL
GRID VALUES
CONVERT TO
MODEL '
FORMAT
HOURLY
GRIDDEO
EMISSIONS
FILE
APPLY
POLLUTANT
PROFILES
Figure 111-2. A Line Source Data Handling Procedure.
-------
are accounted for separately throughout the procedure. Therefore, the hourly
emission factor files contain both an exhaust and an evaporative emission
factor. The result, once the link and Intrazontal emissions are merged, 1s
the Gridded Hourly Emissions File, which contains emissions of exhaust HC,
evaporative HC, NO and CO.
rt
Next, the exhaust HC and evaporative HC are allocated to the five
carbon-bond categories on the basis of user Input data. Also, NO 1s split
A
Into NO and N02. The resulting file, containing all eight pollutants, 1s
then converted to model format.
The hypothetical procedure shown 1n Figure II1-2 1s Intended for
Illustrative purposes. Each transportation agency 1s likely to select
and/or develop Its own procedure and corresponding software. Several
existing network emission models, as the above type of procedure 1s
often termed, are discussed 1n "Volume II. Emission Inventory Require-
ments for Photochemical A1r Quality Simulation Models" (Pacific Environ-
mental Services, 1979).
46
-------
B. AIR QUALITY DATA
In order to simulate the time history of pollutant concentrations,
1t 1s necessary to know or estimate the mass of pollutants entering from
outside the modeling region. These "boundary conditions," expressed as
concentrations, represent the transport of pollution, both ozone and pre-
cursors, Into an urban area from upwind sources. In addition, "Initial
conditions" must be specified as discussed 1n Chapter I, Principles of the
Airshed Model. In other words, the mass of pollutants existing within each
grid cell of the modeling region at the start of the simulation must be known
or estimated, again 1n terms of concentration. Lastly, verification and
evaluation of the model's air quality estimates requires a suitable data
base of measured air quality for comparison. Because existing monitoring
networks are generally not designed to provide boundary conditions, Initial
conditions or adequate verification data, supplementary monitoring 1s usually
required.
1. Requirements for Air Quality Data
As Indicated 1n Table III-4, air quality data are needed to establish
boundary conditions and Initial conditions for model application and to
verify the model results. The pollutants of primary concern are ozone and
Its precursors: nitrogen dioxide, nitric oxide, and nonmethane hydrocarbons.
In addition, the composition of nonmethane hydrocarbons 1n terms of Individ-
ual organic species 1s needed. Data for other pollutants, I.e., carbon
monoxide, are of secondary Importance only.
The monitoring data base should characterize the following aspects of the
photochemical oxldant problem:
47
-------
Table III-4
A1r Quality Data Needs
Boundary Conditions
Ground-level concentrations on
upwind boundary of region, con-
centrations aloft, and temporal
variation of 03, N02, NO and NMHC.
Initial Conditions
Early morning ground-level con-
centrations throughout region,
concentrations aloft, and vertical
profiles of 03, N02, NO, and NMHC.
Verification Data
Hourly ground-level concentrations
of Oo downwind and within urban
area.
Miscellaneous Data
Species composition of NMHC aloft
and on the ground upwind and within
urban area; concentrations of CO.
at existing monitors.
Used to enhance both boundary conditions and Initial condition data.
48
-------
1. Ozone and precursor concentrations transported Into the urban area
from upwind source regions.
2. Precursor concentrations within the urban area as a result of
local emissions.
3. Ozone concentrations within the urban area.
4. Maximum ozone concentrations downwind of the urban area.
The first two aspects need to be characterized 1n order to specify the
Initial and boundary conditions for Input to the model. The third and fourth
provide a basis for evaluating the model's estimates of ozone concentrations.
Ozone concentrations downwind are of concern because this 1s where maximum
ozone levels generally occur. Ozone levels within the urban area are also
of concern because this 1s where population exposures are typically the
greatest. In addition, precursor concentrations within the urban area are
useful for evaluating the model's estimates of hydrocarbon and nitrogen
oxide concentrations.
2. Collection of A1r Quality Data
The data collection program should be designed on the basis of
the upwind, urban, and downwind considerations just described. The siting
of additional monitors 1s almost always required. Before supplemental
monitoring stations are Installed, however, existing stations and Instru-
mentation should be used to best advantage. Relocation of existing sta-
tions should be considered 1f this will Improve the monitoring network.
Typically the data collection effort lasts three months, chosen to coincide
with the "oxldant season." Historical ozone/oxldant data should be con-
sulted, although the highest ozone levels generally occur between June and
September 1n most regions of the United States.
The number and location of upwind (and downwind) monitoring stations
depends on the consistency of the wind flow from any particular sector
49
-------
during high ozone periods. Historical records of air quality and meteor-
ology should be examined to Identify any prevailing wind directions on days
with high measured ozone concentrations. The time of day should be con-
sidered 1n this regard since early morning surface winds tend to be light and
variable whereas mid-day winds are more uniform and better represent the
regional air flow pattern. If a single wind direction quadrant 1s Identified,
one or two upwind monitoring stations will probably suffice; otherwise more may
be required. The monitors should be located fifteen to thirty kilometers
upwind of the edge of the metropolitan area and be sited so as to avoid the
effects of any localized emission sources, Including highways. Ozone, nitric
oxide, nitrogen dioxide, and nonmethane hydrocarbons should be measured at
each upwind station.
Urban monitoring stations should be located so as to characterize the
spatial distribution of ozone and precursors 1n the metropolitan area.
The number of stations depends on the area's size and the configuration of Its
emission sources. At least two stations should be located in the central
business district. Other sites should be selected so as to achieve a
balanced representation of the emissions 1n the metropolitan area and to
monitor areas of special Interest (urban residential, suburban, or Industrial
areas, for example). The total number of urban sites may range from as few
as three to as many as nine. Again, ozone, nitric oxide, nitrogen dioxide,
and nonmethane hydrocarbons should be measured at each urban site.
Additional ozone monitors should be located downwind 1n order to
determine ozone maxima under conditions of downwind transport. Their
number and location depends, as for upwind monitors, on the existence of
a predominant wind direction during periods of high ozone concentrations.
Ozone.maxima frequently occur fifteen to thirty kilometers or more down-
wind of the urban core. The total number of downwind stations may range
50
-------
from as few as four where the wind direction 1s highly persistent to as
many as nine or more where 1t 1s not. At least one station should be
located beyond the expected area of maximum ozone Impact to assure that
the network actually encompasses the highest values. Downwind stations
should be sited away from areas of significant emissions, Including Isolated
point sources and highways. If ozone levels 1n downwind suburban towns or
cities are desired to be measured, additional stations would be needed.
If nitrogen dioxide 1s now or 1s expected to be a problem, additional
Instrumentation at the downwind stations nearest the urban area should be
considered. Some evidence Indicates that maximum nitrogen dioxide levels
may occur five to fifteen kilometers downwind of the urban core.
In addition to the basic monitoring network, Individual organic species
should be measured 1n order to determine the ambient mix of carbon-bonds.
Generally grab samples are taken over a three hour period and then analyzed
* • &
1n the laboratory by gas chromatography for species concentrations. J"he
species data totals also serve as a check on the nonmethane hydrocarbon
measurements. Grab samples should be taken at the urban monitoring stations
/*
and 1f transported hydrocarbons are expected to be significant, at the up-
wind stations as well.
Figure III-3 shows an air quality monitoring network for a major
metropolitan area having a well-defined predominant wind direction from the
west and southwest. In this case, the number of upwind, urban, and downwind
stations 1s 2, 7, and 7, respectively. Ozone 1s measured at all 16 sites
while NOX and NMHC are measured both upwind and within the urban area.
Due to the general urbanization of the area as a whole, NMHC 1s also measured
51
-------
PA
\
ScW •03,NOX \
r _)
,3. NO, BBS"**-*-
--- °3-N
-------
at two near downwind locations while NO and NOg 1s recorded both near and
far downwind. In this example, several other pollutants are measured
which are not always Included 1n a photochemical monitoring program.
These are HN03, PAN, and CO. The nature of the monitoring network for any
particular city will depend on local factors, Including the size of the
urban area, the location of existing monitors, the Importance of Inter-
urban transport, and the prevailing wind dlrectlon(s) during periods of
high ozone concentrations.
Development of a monitoring data base Imposes quality assurance
requirements more stringent than for routine monitoring purposes. Site
exposure should be 1n conformance with established guidelines (Ludwlg and
Shelar, 1978). Routine Instrument checks, Including zero and span, should
be performed twice per week during the three-month study period and Inde-
pendent audits should be performed once a month. These quality assurance
measures should be conducted at both the existing and supplemental monitoring
stations.
When transport of ozone and precursors Into an urban area 1s known
>i'6v
or suspected to be significant, air quality measurements aloft are desirable.
Instrumented aircraft equipped with ozone and nitrogen oxide/n1trie oxide
monitors are generally used. Grab samples should also be collected in order
to characterize the organic species mix aloft. Aircraft measurements are
usually limited to days on which high ozone levels are expected.
C. METEOROLOGICAL DATA
As stated earlier, the Airshed Model simulates the urban atmosphere
under the constraint that pollutant mass 1s conserved. Conservation of
mass requires that pollutant transport through the faces of each grid cell
53
-------
be known or estimated. This Includes both advectlon 1n the horizontal and
dispersion 1n the vertical. Advectlon 1s defined by the time-averaged wind
vector while dispersion 1s related to fluctuations 1n the wind and to vertical
turbulence. Variations 1n the vertical distribution of temperature affect
the vertical extent of mixing within a column of grid cells. The height
to which the mixed layer extends 1s termed the mixing height. In order to
simulate all these processes, a three-dimensional wind field and a two-
dimensional field of mixing heights, as well as atmospheric stability, all
as a function of time, are required.
Another Important meteorological parameter 1s solar Insolation since
several key photochemical reactions are driven by sunlight. The amount of
insolation present, as a function of time, must therefore be known or
estimated.
1. Requirements for Meteorological Data
Meteorological data needed to characterize the wind field, mixing
height, stability, and solar Insolation are listed 1n Table III-5. Hourly
surface observations of wind speed and wind direction are necessary to
describe pollutant transport at the surface while vertical profiles of
speed and direction at various times during the day are necessary to
describe the upper-level winds. Measurements of vertical temperature
gradients at various times are used together with hourly air temperatures
at the surface to estimate hourly mixing heights. Vertical temperature
gradients are also used (along with solar Insolation data) to characterize
atmospheric stability within the mixed layer.
In order to adjust the photochemical rate constants so as to reflect
the amount of radiation available, solar Insolation measurements are
54
-------
Table III-5
Meteorological Data Needs
Wind Field Hourly surface observations of
wind speed and direction through-
out region; vertical profiles of
wind speed and direction and their
temporal variation.
Mixing Height and Stability Hourly surface temperature
observations throughout region;
vertical temperature gradients
and their temporal variation.
Miscellaneous Data Hourly surface observations of
solar Insolation; surface observa-
tions of relative humidity and
pressure and their temporal variation,
55
-------
needed. Surface air temperatures are also used to adjust various chemical
reaction rate constants. Observations of relative humidity are needed 1n
order to compute water vapor concentrations which 1n turn enter Into the
photochemistry of oxldant formation (though the effect 1s minor). Con-
version of pollutant concentrations from a mass to a volume basis requires
that atmospheric pressure be known.
2. Collection of Meteorological Data
The number of meteorological monitoring stations needed Is related
to the size of the city and the complexity of the wind field. Metropolitan
areas located adjacent to coastlines or lakeshores and areas located 1n
complex terrain will require special consideration when designing the
meteorological monitoring program. As discussed 1n Chapter IV, Preparation
of Model Input Data, wind field modeling may be required 1n such situations.
The wind data collection effort would then be directed at satisfying the
needs of the model chosen. Even for Inland areas having relatively flat
terrain, significant heat Island effects may be encountered. Sufficient
spatial resolution 1s required to characterize these effects. Meteorological
data collection should occur over the same three month period as the air
quality monitoring. The following discussion of requirements 1s applicable
to metropolitan areas with generally flat terrain and minimal sea breeze
effects.
In general, wind Instrumentation should be collocated with the air
quality monitoring stations discussed previously. The upwind, urban, and
downwind nature of the regional air flow can therefore be characterized
and upwind-downwind trajectories can be determined. However, care should
be taken to Insure that wind Instruments are sited to avoid mlcroscale
56
-------
wind effects such as may be experienced 1n the lee of buildings or 1n close
proximity to topographic features. Depending on city size and configu-
ration, as well as factors mentioned previously, the minimum required
number of upwind stations may range from one to three, urban stations from
two to five, and downwind stations from two to four 1n number. Wind
Instruments at existing air quality monitoring stations should be
Incorporated 1n the study network 1f possible. National Weather Service
wind data, which 1s usually collected at major airports and should be
available 1n most all metropolitan areas, can be used to further supplement
the wind monitoring network.
Mixing heights are oftentimes determined by measuring the vertical
temperature gradient 1n the atmosphere. Although twice dally mixing heights
are routinely available from the National Weather Service In many metro-
politan areas, the measurements are normally taken 1n rural rather than urban
settings. These data should therefore be supplemented with additional
radiosonde ascents and surface temperature measurements 1n order to charac-
terize mixing heights over the modeling region and their variation with time.
In particular, estimates are needed of the minimum mixing height before
sunrise, the mixing height during the nrld-morning when the rate of rise 1s
quite rapid, and the maximum mixing height 1n the afternoon. A minimum of
three ascents each day should be made at an urban site with release times
of 0400, 1000, and 1500, for example. Upwind of the city, the primary con-
cern Is the development of the mixing layer after sunrise. Therefore,
radiosonde releases should also be conducted at an upwind site at 0500
and 0900, for example. Alternatively, an acoustic sounder may be an
advantageous and economical method of obtaining morning mixing heights at
57
-------
the upwind site. (However, if an acoustic sounder 1s used, an additional
plbal site should be added to those suggested below.)
If aircraft flights are conducted to measure pollutant concentrations
aloft, the aircraft should be equipped with temperature sensors to further
t
define the temperature structure of the atmosphere. In addition, since
surface temperature measurements CA" be used to help Interpolate mixing
heights temporally and spatially, such measurements should be taken at the
wind monitoring stations discussed previously.
The radiosonde ascents should be supplemented with plbal ascents 1n
order to characterize air flow aloft. By tracking the plbal or radiosonde
over time, vertical profiles of wind speed and wind direction can be
Inferred. Generally two or three plbal release sites are desirable.
The locations and times of their release should be coordinated with the
radiosonde ascents so as to provide a maximum overall representation of
the regional wind field.
Solar Insolation (both direct and diffuse) 1s readily measured using a
pyranometer. The number of sites depends on the size of the city, and
the presence of water or terrain which might Influence cloud cover. In
general however, two monitoring sites are desirable. The Instruments should
be collocated with the wind and air quality Instrumentation at stations
representative of upwind, urban, or downwind locations.
In regard to relative humidity and atmospheric pressure, no additional
data need normally be collected. National Weather Service or local agency
surface measurements, normally taken every hour, are generally sufficient for
modeling purposes.
Figure III-4 shows a meteorological monitoring network for a major
metropolitan area having a well-defined predominant wind direction from
58
-------
W8
WD
SR
T
RW
PB
Wind Speed
Wind Direction
Solar Radiation
Temperature
Rawlnnnde
Pibal
ATLANTIC
OCEAN
Figure 111-4. A Meteorological Monitoring Network for Philadelphia.
59
-------
the west and southwest during periods of high ozone concentrations. In
this example, the number of upwind, urban, and downwind surface sites 1s
2, 5, and 5 respectively. Wind speed and wind direction are measured
at each station while temperature 1s recorded at selected sites upwind,
within, and downwind of the urban area. Solar radiation 1s measured
both upwind and downwind. Winds aloft are measured at three sites, one
each at an upwind, urban, and downwind location* Temperature soundings
are also taken at the urban site. The number of meteorological monitoring
sites for any particular city will depend on the nature and complexity
of the wind field and of the surrounding terrain.
60
-------
CHAPTER IV. PREPARATION OF MODEL INPUT DATA
Once the emissions, air quality, and meteorological data described
in Chapter III have been assembled, the data must be prepared for model
input. Data preparation includes selection of a set of individual modeling
days from the meteorological and air quality data bases for which model
input data is to be prepared. Once the data has been prepared, the model
can be run. Model operation is described in detail in the User's Manual
u
for the Airshed Model (Ames, et. al., 1978) and will not be discussed here.
Computer resources necessary to run the model are, however, described
briefly in Chapter VI, Resource Requirements for an Airshed Model Appli-
cation.
A. DAY SELECTION CRITERIA
Prior to attempting to prepare the aerometric data for model input, one
must select the days to be modeled. Normally, three or four days should be
selected. Several selection criteria should be observed. First, ozone
levels should be near the maximum that occurred during the air quality
monitoring program. Such high ozone episodes often occur over a two or
three-day period. Second, weekend days should be avoided. Use of an
emission Inventory developed for a weekday may Introduce errors if used on a
weekend day. Total regional emissions are lower on weekends and the temporal
and spatial distribution of the emissions are different. Third, the
At this time, only a draft has been provided by Systems Applications,
Inc. The manual will be available for public release upon completion of a
finalized version.
61
-------
aerometHc data base should be relatively complete. Excessive occurrences
of missing data makes data preparation difficult and may lead to a poor or
suspect verification. Fourth, the air quality and meteorological data should
not exhibit abnormal complexity. Unusual wind fields or concentration
gradients may place undue stress on the model. Such occurrences may cause
the data itself to be suspect in some cases. Passage of warm or cold fronts
may introduce unacceptable complexities. Under no circumstances should any
precipitation have occurred during the pre-dawn to mid-afternoon hours. In
general, sunny skies and light winds associated with slow-moving high pressure
systems are conducive to high ozone levels.
B. DATA INPUT FILES
the Airshed Model contains eleven data preparation programs which assist
the user in creating the aerometric and emissions input files required for
a simulation run. Figure IV-1 shows the input data files while Table IV-1
provides a short description of each. A variety of alternate methods for
preparing the aerometric data are possible. Selection of a method depends
on the nature of the data base. The level of sophistication of the method
selected follows from the level of sophistication of the data base. Several
methods depend upon Interpolation/extrapolation routines 1n the horizontal
using station measurements. Vertical resolution 1s generally specified by
means of normalized vertical profiles. Different methods may be used for
preparing the data for different subregions. This allows the modeler to
better utilize a spatially uneven data base as well as characterize spatial
differences more realistically. Table IV-2 Indicates the spatial and
temporal resolution for each data input parameter that is achieved by
utilizing the preparation programs.
62
-------
METEOROLOGICAL DATA
AIR QUALITY DATA
ri
WIND
DIFFBREAK
ri
METSCALARS
CO
REGION TOP
AIR QUALITY
TEMPERATUR
TERRAIN
BOUNDARY '
TOPCONC
I
AIRSHED
SIMULATION
PROGRAM
Figure IV-1. Input data files for airshed simulation program.
EMISSIONS DATA
ri
EMISSIONS
PTSOURCE
-------
Table IV-1.
DESCRIPTION OF THE INPUT FILES TO THE AIRSHED MODEL
DIFFBREAK This file contains the mixing height for each column
of cells at the beginning and end of each hour of the
simulation.
*
REGIONTOP This file contains the height of each column of cells at
the beginning and end of each hour of the simulation. If
this height Is greater than the mixing height, the cell or
cells above the mixing height are assumed to be within an
Inversion.
WIND' This file contains the x and y components of the wind
velocity for every grid cell for each hour of the simu-
lation. Also the maximum wind speed for the entire grid
and average wind speeds at each boundary for each hour
are Included on this file.
METSCALARS This file contains the hourly values of the meteorological
parameters that do not vary spatially. These scalars are
the N02 photolysis rate constant, the concentration of
water vapor, the temperature gradient above and below the
Inversion base, the atmospheric pressure, and the exposure
class.
This file contains the Initial concentrations of each
species for each grid cell at the start of the simulation.
\
This file contains the location of the modeling region
boundaries! This file also contains the concentration of
each species that 1s used as the boundary condition along
each boundary segment at each vertical level.
TOPCONC This file contains the concentration of each species for
the area above the modeling region. These concentrations
are the boundary conditions for vertical Integration.
TEMPERATUR This file contains the hourly temperature for each surface
layer grid cell.
•
EMISSIONS This file contains the ground level emissions of NO, N02,
five carbon bond categories, and CO for each grid square
for each hour of the simulation.
AIRQUALITY
BOUNDARY
64
-------
Table IV-1. (Continued)
PTSOURCE This file contains the point source Information,
Including the stack height, temperature and flow
rate, the plume rise, the grid cell Into which the
emissions are emitted, and the emissions rates for
NO, N02, five carbon bond categories, and CO for
each point source for each hour.
TERRAIN This file contains the value of the surface roughness
and deposition factor for each grid square.
65
-------
Table IV-2
DATA-RELATED INPUT PARAMETERS FOR THE AIRSHED MODEL
Description
METEOROLOGY
Horizontal (u-v)
winds (m/sec)
Reference height of
surface wind monitoring
stations (m)
Diffusion break (m)
Spatial and
Temporal
Resolution*
XJJZ t
Top of modeling x
region (m)
Ground-level tempera- x
tures (°C)
Surface atmospheric
pressure (mb)
Temperature gradient below
diffusion break (°C/m)
Temperature gradient above
diffusion break (°C/m)
Water concentration 1n
the atmosphere (ppm)
>
Exposure (stability class)
Radiation Intensity
factor (per m1n)
Surface roughness (cm) x
Remarks
The vertical component, w, 1s computed
by the Airshed Model, rendering the
resultant wind field mass consistent
Used 1n the d1ffus1v1ty algorithm
x Elevation at which the stability
structure of the atmosphere changes
markedly (e.g., an Inversion or
thermal Internal boundary layer)
x Generally defined 1n relation to
diffusion break
x Used 1n kinetic module
x .Used 1n computing grid cell concen-
trations associated with emissions
x Used in plume rise calculations
x Used In plume rise calculations
x Used 1n kinetic module
x Uspd 1n dlffuslvlty algorithm
x Used 1n kinetic module
Used in dlffuslvlty and surface
sink algorithms
66
-------
Table IV-2 (continued)
Description
Vegetation factor
AIR QUALITY
Initial conditions
(pphm)
Boundary conditions
(pphm)
Concentrations above
top of modeling
region (pphm)
EMISSIONS
Lumped ground-level
emissions from
stationary and mobile
sources (gm-moles/hr)
Elevated stationary
point source emissions
(gm-moles/hr)
Elevated point source
stack data Including
height (m) temperature
(°K) diameter (m) and
velocity (m/sec)
Spatial and
Temporal
Resolution*
x x
x x
x x
Remarks
Used 1n surface sink algorithms
Required for NO, N02, 0,, five carbon
bond categories, PAN, BZA, and CO
Required for same pollutants as
above
Required for same pollutants as
above
Required for NO, N02, five carbon-
bond categories and CO
Emissions from tall stacks for
the above pollutants are
required
Used in plume rise calculations
xy Indicates two dimensional data field
xyz Indicates three ^Pfinensiohal data field
t indicates time varying data field
67
-------
1. Preparing the Meteorological Data
Preparation of the METSCALAR and TEMPERATUR input files is relatively
straightforward. The METSCALARS file contains those meteorological parameters
which do not vary spatially. Hourly values of the average temperature gradi-
ent below and above the mixing height and hourly values of atmospheric pres-
sure, water concentration (derived from humidity measurements), and exposure
(stability) class must be specified. In addition, hourly N02 photolysis rate
constants must be input. The latter are readily calculated using measured
solar insolation and estimated solar angle (Schere and Demerjian, 1978).
Preparation of the TEMPERATUR file requires that hourly station temperature
readings must be input and the desired interpolation routine selected.
Preparation of the DIFFBREAK file generally requires the use of
algorithms or techniques external to the data preparation program. Vertical
temperature gradient readings (2 to 3 sites, 3 to 4 times daily) and surface
temperature readings (5 to 12 sites, hourly) are used to determine hourly
mixing heights at all stations (7 to 15 sites). These values are then input
to the data preparation program and a suitable spatial interpolation routine
is selected. Creation of the REGIONTOP file does not involve any actual
data input. Rather, the user generally specifies the number and height of
grid cells above the mixing height. Alternatively, a fixed region height
may be used.
Preparation of the WIND file requires special expertise in wind
field modeling and analysis and the use of specialized wind generation
programs. Such programs are not provided by the Airshed Model. The WIND
file should be recognized as a critical element of satisfactory model per-
formance. For these reasons, the WIND file is undoubtedly the most
68
-------
challenging file to prepare. An Iterative procedure, whereby wind field
modeling 1s conducted and the resulting wind field 1s evaluated, 1s usually
necessary.
Three distinct approaches toward generating wind field Inputs may be
taken. The simplest approach Is the use of an Interpolation scheme whereby
station readings of wind speed and wind direction are used to approximate
values 1n each grid cell. Vertical wind profiles derived from plbal,
rawlnsonde, or other measurements, are also used. The Interpolation scheme
must be followed by a smoothing of the resulting multilayer, horizontal
wind fields. This 1s necessary to reduce divergence in the resulting wind
field. The Airshed Model simulation program Itself then removes any remain-
Ing divergence by generating compensating vertical winds, thus Insuring a mass
consistent wind field.
An alternative and more sophisticated approach 1s the use of so-called
"diagnostic models." Such models are appropriate where the local terrain con-
figuration has a significant effect on wind patterns. These models, which
Incorporate some of the same concepts discussed above, use a basic conserva-
tion of mass equation subject to various boundary constraints. Such con-
straints may Include heat Island effects as well as terrain blocking and
channeling effects and Inversion limiting effects.
The most complex approach 1s a dynamic wind field model based on the
solution of mass, momentum, and energy equations. These models may be needed
In areas with significant thermally generated sea or lake breezes. Unfortu-
nately, their data Input requirements are considerable and their computer
requirements are comparable with those of the Airshed Model Itself.
A technique, which 1s available as an Airshed Model data preparation
program, may be used in special situations where terrain and water effects
69
-------
are negligible and little wind data 1s available. It uses an average wind
vector together with a network of station temperature readings to approximate
urban heat island effects.
Preparation of the TERRAIN file involves the estimation of surface
roughness and surface deposition factors throughout the modeling region.
Pollutant uptake by water bodies and vegetative surfaces may influence
ozone concentrations. Standard factors are used to represent uptake by
various types of surfaces such as trees, grasses, water, and man-made mate-
rials. The relative composition of individual grid squares and/or subregional
areas in terms of these surface types can be approximated using land-use maps
and assumed mixes for each land use category. Surface roughness 1s employed
by the model to characterize vertical dispersion near the ground. Land-use
maps, topographic maps, and the surface types described above are used to
determine the roughness controlling features so standard roughness length
values can be selected. As with surface deposition, roughness lengths may
be specified at the subregional level or at the grid square level depending
on Its spatial variability.
2. Preparing the A1r Quality Data
Three files use air quality data: BOUNDARY. AIRQUALITY, and TOPCONC.
Although each 1s created by a separate data preparation program, care should
be exercised to assure the files are mutually consistent. For example, the
vertical distribution of upwind boundary conditions should be consistent with
the values used above the modeling region. Otherwise, an artificial disconti-
nuity is introduced. Each file contains values for the eleven pollutants
listed in Table IV-3.
70
-------
Table IV-3
POLLUTANTS TREATED BY THE AIRSHED MODEL
Pollutant Airshed Model Name
Ozone 03
Nitric Oxide NO
Nitrogen Dioxide N02
Organlcs:
Highly Reactive Double Bonds OLE
Aromatic Rings ARO
Single-Bonded Carbon Atoms PAR
Carbonyl Bonds CARB
Moderately Reactive Dbuble Bonds ETH
Carbon Monoxide CO
Benzaldehyde BZA
Peroxyacetyl Nitrate PAN
71
-------
The AIRQUALITY file contains the initial concentration of each pollutant
1n each grid cell at the start of the simulation. Prior to use of the prepara-
ration program, nonmethane hydrocarbon values at the upwind and urban sta-
tions for the first hour are converted Into the five carbon bond categories
using the organic species data from early morning grab samples. (This con-
version step 1s discussed further in regard to emissions in the Appendix,
Technical Description of the Airshed Model.) Then these values, along with
values of ozone, nitric oxide, and nitrogen dioxide are specified for each
station at which they were measured. A suitable (horizontal) Interpolation
routine 1s then selected, one for each pollutant. If aircraft spirals were
used to collect upper-level early morning air quality data, a vertical pro-
file for each pollutant and the location the spiral was flown 1s specified.
If not, a simple vertical profile must nevertheless by chosen. For example,
values at the ground might be extended up to the mixing height. Estimated
values aloft (discussed below) could be used above the mixing height. When
Interpolating horizontally or specifying a vertical profile, the modeling
region can be broken up Into subreglons. However, designation of subregions
must be observed for all pollutants.
For carbon monoxide interpolation routines may be used with one-hour
data from the existing monitoring network. Without data aloft, values
could be assumed constant 1n the vertical. For those pollutants not mea-
sured during the monitoring program (BZA and PAN), typical reference values
may be used (Ames, et. al., 1978).
The BOUNDARY file contains the hourly concentrations of each pollutant
for each boundary cell. The region boundaries must be defined first, how-
ever. This 1s done by assigning Hne segments which completely enclose the
72
-------
region. A ground-level value for each pollutant 1s then Input for each
boundary segment each hour. Deriving these values from the air quality data
base 1s mostly a matter of judgment. The prevailing wind direction each hour
1s used to determine which stations are upwind or downwind. Downwind boundary
values are not Important although they must be input. As for Initial con-
ditions, nonmethane hydrocarbons must be converted to five carbon bond
categories. A normalized vertical profile must also be specified for each
boundary segment for each pollutant. These can be varied through the day.
The vertical profiles should be tied to the ground-level value and the value
aloft. If aircraft data has been collected, profile preparation Involves
selecting suitable profile methods and preparing the profile data sets so as
to make the best use of available data. Even 1f upper-level data have not
been collected, a simple assumed profile must still be entered.
The TOPCONC file contains the hourly concentrations of each pollutant
above the modeling region. If aircraft data have been collected the data
are used to determine the concentrations aloft of ozone, nitrogen dioxide,
and nitric oxide. The results of aircraft organic species grab samples are
used to estimate the concentrations of the five carbon-bond categories. If the
data show any significant spatial variation, this should be represented either
by using an Interpolation routine with so-called "station" readings or by
specifying a value for each grid square across the top of the region. Sig-
nificant temporal variations should also be taken Into account. If no upper-
level air quality data are available, values aloft may be estimated crudely
by using the late morning air quality measurements at an upwind ground-level
monitoring station. This method assumes that upwind ground-level concentra-
tions are Indicative of concentrations aloft upon dissipation of the nocturnal
73
-------
surface Inversion. For those pollutants not measured either upwind or aloft
during the monitoring program (CO, BZA, or PAN) typical reference values may
be used (Ames, et. a., 1978).
3. Preparing the Emissions Data
As discussed 1n Chapter III, Data Needs, the emissions data base
should consist of hourly grldded ground-level emissions from minor point sources,
area sources, and mobile sources. Each of these three files should contain the
emissions for eight pollutants: NO, N02, OLE, PAR, CARB, ARO, ETH, and CO.
These three files must then be merged to create the EMISSIONS file required
by the Airshed Simulation Program.
While the EMISSIONS file contains the ground-level emissions, the
elevated point source emissions are contained on the PTSOURCE file. Prepara-
tion of this file involves specifying the location and stack parameters
(height, diameter, velocity, and temperature) for each major point source,
entering the hourly emissions of the eight pollutants for each source, and
running the data preparation program provided. The data preparation program
utilizes the meteorological Input files described earlier to compute plume
rise and then assign the point source emissions to individual elevated
grid cells.
74
-------
CHAPTER V. EVALUATION OF MODEL PERFORMANCE
Before a model 1s used for assessing alternate control measures for
reducing ozone precursor emissions, one must demonstrate the ability of
the model to estimate ozone concentrations satisfactorily. Such a veri-
fication on several high ozone days 1s desirable. By simulating several
meteorological and air quality situations* model performance can be
evaluated over a range of conditions conducive to ozone formation. Should
verification attempts fall on one day, other days are stm available for
analysis. Also, control measures may be more or less effective depending
on the level of pollutant transport and the mesoscale meteorological
conditions. In addition, verification on multiple days allows control
strategies to be tested under different conditions.
Normally, single day verifications should be conducted for at least 14
hours, usually from about 5:00 a.m. to about 7:00 p.m. Such a time span
covers the morning emissions peak and the afternoon ozone peak should it
occur late 1n the day. However, 1f downwind ozone monitors clearly show that
ozone levels drop off much earlier 1n the day, the length of the simulation
may be shortened.
A number of performance measures may be used to evaluate whether or not
the model 1s performing satisfactorily. Normally, performance 1s evaluated
separately on each verification day. Both graphical and statistical tech-
niques are useful. Several techniques should be chosen because different ones
measure different aspects of the model's performance. Some measure the model's
performance at estimating peak concentrations, others measure spatial and temporal
correlations, while still others measure the overall accuracy of the model results.
The U. S. Environmental Protection Agency has recently published two
reports addressing the evaluation of model performance: Procedures for
75
-------
Evaluating the Performance of Air Quality Simulation Models (Hlllyer,
Reynolds, and Roth, 1979) and Performance Measures and Standards for Air
Quality Simulation Models (Hayes, 1979). The first report summarizes
various issues and considerations pertaining to model evaluation studies.
The second report, in part, presents a number of specific performance measures.
Comparisons of model-estimated concentrations and observed concentrations
must be done cautiously. Obviously, Instrument error 1s always a complicating
factor. Beyond this, however, the observed value at best represents a con-
centration measured at a single point. In contrast, the estimated value
represents an average concentration within a surface layer grid cell; gradients
within a cell cannot be accounted for. The Airshed Model does perform an
east-west and a north-south linear interpolation between the four closest
grid cells to obtain a point estimate. The grid cell average value 1s used
to represent the value at the centroid of the cell for the interpolation.
This procedure does not, however, compensate for the fundamental difference
between observed and estimated values. It only provides an objective
method for comparing model results to measured concentrations. Clearly, the
representativeness of the monitor site is a key factor. For example, model
estimates of NO or CO that are much lower than monitor readings while at
the same time ozone estimates that are significantly higher may indicate the
monitor is being unduly affected by localized emission sources under certain
meteorological conditions. These kinds of disparities point up the need for
properly siting ozone monitoring equipment when setting up the ambient
network discussed in Chapter III. In general, the observed concentrations
are nevertheless assumed to be the "true values" when using performance
measures to evaluate model results.
76
-------
The following discussion presents a variety of performance measures,
both graphical and statistical, which have been Identified for use in
evaluating model performance. The measures discussed here are not Intended
as recommendations, as further experience with a wide variety of measures
in various applications 1s needed in this newly developing field.
A. GRAPHICAL TECHNIQUES
Scatter diagrams are one of the most common graphical methods of com-
paring observed and estimated concentrations. Normally the observed values are
plotted on the abscissa on the same scale as the model estimates which are
plotted on the ordlnate. A 45 degree 11ne through the origin then represents
a perfect correlation. The distance of points from this line 1s a measure of
the deviation while the clustering of points above or below is a measure of
bias. As shown in Figure V-1 maximum values at each station, mean values
over a certain time Interval (or the entire simulation), or all hourly values
at all stations may be visually compared in this way.
Time histories of observed versus estimated concentrations at individ-
ual stations may also be presented graphically as shown in Figure V-2.
Such plots show whether the model estimates conform to the temporal pat-
terns exhibited by the monitoring data and the extent to which model esti-
mates are out of phase with actual values.
If the monitoring network is sufficiently dense, it may be possible to
plot isopleths of the observed concentrations and compare these to isopleths
of the model estimates, plotted at the same scale, as shown in Figure V-3.
Two attributes may be examined using such plots: general pattern corres-
pondence and directional alignment. General pattern correspondence is more
important than simple directional alignment. Nonalignment may be caused
77
-------
9- 5
a.
0 5 10 15
Observed concentrations (pphm)
14
1.
1
.12
_ b.
i i i I r
DSL.
DRCO
DBUo
I ! I I I
oOLI -
DRMo.
DSFo
2345 67
Observed mean (pphm)
i8
1
a.
i
1
I
4 8 12 16
Observed maximum (pphm)
Figure V-l Scatter diagrams of (a) observed oxidant concentrations versus
calculated hourly average ozone concentrations, (b) observed versus calculated
station mean concentrations, and (c) observed versus calculated station maximum
concentrations (Duewer, MacCracken, and Walton, 1978).
78
-------
-OBSERVED
—-«-fltEO!CTED
»-0.—A—.0 PARKER RO
I f i f ft ff tt ¥ 1 I i i t »
T
-------
00
o
ESTIMATED
OBSERVED
Figure V- 31. Comparison of estimated and observed ozone isopleths for the hour
of maximum observed concentration.
-------
by small errors 1n the wind field while a lack of pattern correspondence
Indicates a more fundamental problem with the data Inputs or the model
Itself. Plots such as Figure V-3 may be further Interpreted by comparing
the areas over which ozone levels exceed certain threshold values above the
standard.
B. STATISTICAL TECHNIQUES
As shown 1n Table V-l, a variety of numerical and statistical
measures may be used for evaluating model performance. It 1s Important to
note at the outset that 1n most applications, the number of days modeled 1s
Insufficient to make any statistical statement regarding the ability of the
model to estimate peak regional oxldant concentrations. Therefore 1t 1s not
possible to say how the model might be expected to perform on a day other
than those for which a verification analysis 1s actually undertaken.
The ability of the model to accurately estimate peak ozone concen-
trations 1s crucial to the application of the model for SIP planning pur-
poses. The most obvious numerical measure 1s the difference between the
estimated maximum ozone concentration at any station and the observed
maximum ozone concentration at any station. This comparison 1s made Irre-
spective of the time of day during which the maxima occurred. Another
numerical measure of peak performance 1s derived by taking the ratio of
the estimated maximum to the observed maximum concentration, one ratio for
each station and then finding the median ratio. The median rather than
the average value is used to avoid unnecessary bias that might be caused
by a bad estimate at one or two stations, particularly at locations where
both values are low.
81
-------
Table V-1. Some Numerical and Statistical Performance Measures
Performance Attribute Performance Measures
Accuracy of Peak
Estimates
Overall Accuracy
of Estimates
Absence of Bias
Diurnal Pattern
Spatial Pattern
Difference between estimated and observed
maximum concentration regardless of time
or location
Median ratio of estimated to observed
maximum station concentrations
Median relative absolute deviation between
estimated and observed maximum station
concentrations
Correlation coefficient of estimated and
observed maximum station concentrations
Median relative absolute deviation between
estimated and observed station concen-
trations taken over all stations and all
hours
Average absolute deviation and standard
deviation of absolute deviations between
estimated and observed station concen-
trations taken over all stations and all
hours
Difference 1n the frequency of occurrence
of values above some threshold, between
estimated and observed station concentra-
tions considering all stations and all hours
Correlation coefficient of estimated and
observed station concentrations taken over
all stations and all hours
Relative deviation between estimated and
observed maximum concentrations averaged
over all stations
Relative deviation between estimated and
observed station concentrations averaged
over all stations for hours when the
observed value exceeds some threshold
Temporal correlation coefficient qf
estimated and observed station concentra-
tions averaged over all stations
Spatial correlation coefficient of
estimated and observed station concentra-
tions averaged over all hours
82
-------
The median relative absolute deviation of maximum values, taken over
all stations, times 100, 1s a good measure of the percent error associated
with the peak ozone model estimate. The relative absolute deviation
1s defined here as the absolute value of the difference between the esti-
mated and observed values, divided by the observed value. Another good
statistical measure of peak model performance 1s the correlation coefficient,
taken over all estimated/observed maximum concentration pairs, one pair for
each station.
Because the Intent of control strategy analysis 1s to reduce ozone
concentrations below the standard and to reduce population exposure, it 1s
also Important that the model perform satisfactorily over a wide range of
ozone concentrations. As with maximum concentrations, the median relative
absolute deviation 1s also a measure of the overall percent error of the
model estimates. In this case, however, rather than limiting the calculation
to the maximum values, the median 1s taken using the entire set of deviations
for all hours during the simulation. The overall correlation of all esti-
mated and observed concentration pairs may also be computed. Measures of
the absolute error and Its variability are the average absolute deviation
and the standard deviation of absolute deviations, respectively. The latter
statistic Involves the computation of the deviation of the absolute deviation
of an estimated/observed pair from the average absolute deviation of all
such pairs. Both statistics are computed over all stations and over all
hours of the simulation.
A somewhat different performance measure Involves the comparison of
the cumulative frequency distribution of estimated ozone concentrations
to the cumulative frequency distribution of observed ozone concentrations,
83
-------
again taken over all stations and all hours as 1n Figure V-4. A numeric
comparison 1s then made of the frequency with which various concentration
levels are exceeded.
One should also evaluate the model results for systematic bias. Bias
occurs when the model has a tendency toward repeated errors 1n the same
direction, either over-estimation or underestimation. In this case, the
algebraic average deviation rather than the absolute average 1s used.
Normalizing the deviation by the observed concentration before averaging
gives the average relative deviation, a measure on a fractional basis of
the seriousness of the bias over any range of concentrations. Possible
variations of this measure Include restricting the computation to the maximum
values for each station or to the hours during which the observed concen-
tration exceeds some value, such as the ambient air quality standard. The
overall bias may be assessed by extending the computation to Include all
stations and all simulation hours.
Quantitative statistical measures may also be used to evaluate the
model's ability to generate satisfactory diurnal and spatial patterns. The
temporal correlation coefficient, averaged over all stations, and the spatial
correlation coefficient, averaged over all hours, are two such Indices.
Variations Involve stratifying the data so that, for example, the temporal
correlation 1s determined for stations Individually or for groups of sta-
tions. Similarly, spatial correlations may be taken over one or more hours,
usually the peak period.
C. REVERIFICATION
In the event of poor model performance, one must determine what
1s causing the model estimates to be unsatisfactory. The possible causes
84
-------
00
01
40
30
20
S 10
5 9
£ 8
g 7
2 6
111 I I I I I—I I I I I I I II I I I I
- -O- — OBSERVED
ESTIMATED
279 DATA PAIRS FROM 3 DAYS. 14 HOURS, 9 STATIONS
I I I I I I I V 1 I I I I I I LJ I I I I I
0.1 0.2 12 5 10 20 30 40 50 60 70 80 90 95 98 99 99.5 99.8 99.9
'' CUMULATIVE FREQUENCY
Figure V-4. Cumulative frequency distribution of estimated and observed ozone concentrations (adapted
from Hayes, 1979).
-------
may be grouped Into three areas: (1) the emissions, air quality, or mete-
orological data bases may not adequately or properly characterize conditions
on the day modeled; (2) the methods used for preparing the air quality and
meteorological data for model Input may not be appropriate or were exercised
Improperly; (3) the model may not be suitable or does not simulate the
atmospheric phenomena Important for the day or location being modeled. Unless
a research and development type effort 1s being contemplated, one should
probably not subject the last cause to any prolonged Investigation. Rather,
efforts should be placed on Investigating the data base and Improving and/or
correcting the methods by which the data base was prepared for model Input.
Generally, Investigation of model performance problems should center
Initially on the procedures used for preparing the aerometrlc data for model
Input. This Is because the modeler has the most control over this portion of
the modeling study. Invariably, deficiencies exist 1n the data base which
necessitate assumptions 1n order to prepare the data for the model. Such
assumptions should be reevaluated. For example, preparation of the wind field
and mixing height data are prime candidates for review. These two Inputs are
*,,
primary driving forces in the model simulation while the number of data points
available and the methods used for their preparation methods may be less than
Ideal. Further analysis of the Input data may Improve model performance.
When Investigating the data base, attention 1s most often focused
on the emission Inventory. This 1s due to the fact that emission Inventories
are frequently found to be deficient 1n one or more respects. One may find
for example, that significant sources of hydrocarbons have been overlooked.
One may also find that not all the vehicle traffic has been accounted for
or that the Input data used to compute mobile source emission factors 1s
86
-------
not representative or sufficiently resolved spatially and temporally.
Another candidate for Investigation Is the organic species data used for
allocating hydrocarbon emissions to the five carbon-bond categories. One
might find for example that serious errors 1n judgment have been made 1n
applying an Imperfect organic species data base to specific sources 1n
the Inventory.
The modeler, due to his familiarity with the data base, the various
methodologies, and their strengths and weaknesses should have a good Idea
of what areas warrant Investigation and modification. Nevertheless, a
>
knowledge of the sensitivity of the model to changes in Input data 1s
Invaluable. For example, little 1s gained from expending much effort on
Devaluating stability class data If the model results are not sensitive
to stability class. While no comprehensive sensitivity study of the Airshed
Model has been reported, Table V-2 summarizes the results of a number of
sensitivity analyses performed with previous versions of the model and
another photochemical grid model, the Llvermore Regional A1r Quality (LIRAQ)
model. If all reasonable and objective modifications to the Input data fall
to upgrade model performance to a satisfactory level, then one 1s forced to
admit that the model 1s unable to simulate the atmospheric conditions existing
on that particular modeling day. In such cases, 1t 1s especially Important
that modeling of several different high ozone days has been provided for
previously.
In summary then, model verification 1s not in any way a "push-button"
operation. Rather, it 1s an Iterative and an Incremental process. With
each execution of the simulation program model performance 1s evaluated,
problems are diagnosed, the Input data 1s reanalyzed, and the Input files are
modified until a satisfactory verification 1s achieved.
87
-------
thrill, £; *TnnS+Try f ?e;?1tivi;^ of Selected Photochemical Grid
Models to Input Data Variations (Reynolds, Tesche, and Reid, 1978)
Study Group
Nodel Version
anA Attributes
HacCracken and
Sauter (I97S)
00
00
Deaerjlan (1976)
Liu tt •!. (1976a)
LIRAQ photochemical model
Two-dimensional time-
dependent grid Mxfel
LiMped kinetic mech-
anism siMilar to
Hecht-Selnfeld-Dodge
Mechanism
Mass-conserving Mind
field
SAI photochemical codel:
1973 version
SAI photochemical Mdel:
1973 version
25 x 25 x 6 grid
IS step Hecht-Selnfeld-
Dodge kinetics
Price numerical Method
Empirical diffusion
algorithm
Two-dimensional Mind
field
Two-dimensional
Initial conditions
Sensitivity Analysis
Variations
Relative humidity Mas
reduced fro» 40* by 20X
Nominal temperature Mas
Increased fro* 285°* to
304*K
Light Intensity Mas re-
duced by SOI
Light Intensity Mas
Increased by a factor
of 2
Initial hydrocarbons Mere
Increased by a factor of 2
Initial NO; concentrations
were Increased by a factor
of 2
Boundary conditions were
reduced by SOX
Initial and boundary con-
ditions were reduced by
SOX
Wind directions were
randmly oerturbed by
0 or *22.5«
Hind speeds wen randomly
perturbed by 0 or ±1 mpti
Hind station Measure-
ments were:
Increased SOS
Increased 2SX
Decreased 25X
D*er««ad SOS
Influence on Model Predictions
Peak ozone Increased by 31
and peak NO; decreased by «J
Peak ozone decreased by 21.
and peak NO; Increased by 5X
Peak ozone decreased by 70S.
and NO; peak Magnitude rewined
unchanged but was delayed 4
hours
Peak ozone Increased by 100S.
and NO; peak Magnitude slight-
ly Increased and preceded base
case peak by 1-3/4 hours
NO; peak Increased by 61 and
was delayed approximately I
hour; ozone peak was not re-
ported, but the Increase in
ozone concentrations was delayed
by up to 3 hours
NO; peak Increased by 101 and
was delayed slightly! 03 re-
Mined unchanged
•Minor" differences occurred
In ozone prediction In the
eastern and northern portions
of the L.A. basin; "significant"
differences were observed In the
western and central portions of
the basin
Predicted ozone In the northern
and eastern edges of basin were
reduced 20 to 30X
A 6.9S average deviation for
•mually prepared and 4.91
for automatically prepared
wind fields (based on CO
predictions) resulted
A 4.91 average deviation for Man-
ually prepared and 2.6X for auto-
matically prepared Mind fields
(based on CO predictions) resulted
HaxlMUM absolute deviation from
the base case results for CO
were-.
19.61
11.BX
20.2%
Sl.TJ
Remarks
LIRAQ sensitivity runs focused
on the kinetic module; accord-
ingly, sensitivity results are
more reflective of SMDQ charter
simulations than urban airshed
simulations.
In the automatic Mind field
studies, perturbations were
•ade to the Monitoring station
measurements and then automatic
procedures were employed to
derive gridded wind fields. In
the Manual wind field cases.
perturbations were Made to the
gridded wind fields after they
had been prepared Manually.
The response of the Model to
variations In wind speed varies
with each chemical species and
is time dependent.
-------
Table *V-2 (Continued)
Study Croup
Model Version
and. Attributes
Sensitivity Analysis
Variations
Horizontal diffusion MS
decreased to 0 and In-
creased to 500 mVs
Vertical dlffuslvlty MS
decreased to 0.5 m'/ssc
and Increased to 50 mz/sec
Nixing depths were In-
creased and decreased by
251
oo
vo
Radiation Intensity MS
Increased and decreased
by 30%
(Missions rate (ground
based) MS Increased and
decreased by 151
Influence on Model Predictions
For Ku • 0. the Maxima abso-
lute deviation for CO ranged
between 0.52 and 2.021 fron.
0600 to 1600 hours
For KH • 500 m2/sec. the
maximum absolute deviation for
CO ranged between 4.4 and 12.9%
from 0600 to 1600 hours
The effect of varying vertical
dlffuslvlty by an order of Mag-
nitude MS about the same as
that of varying the wind speed
by 25 to 50*
MaximiM absolute percentage
deviations for the Increase and
decrease, respectively. Mere:
For CO. Stand 121
For HO. 11% and 18.51
For N02. 8.51 and 15.51
For 03. 11.51 and 231
Remarks
Haxliwm absolute percentage
deviations for the increase
and decrease, respectively.
For NO. 17land 401
For N02. 741 and 551
For 03. 91 and 1)1
The effects of Increaslno
and decreasing emissions
rates are alMOst Identical;
peak basin-wide average per-
centage changes In CO and
NO? nere about the sane
(6-81)
The base case value MS S
The buildup of the Mixing depth
variation effect is time
dependent.
Decreasing the Mixing depth has
a greater effect on the ground-
level concentrations than 1n-
creasinq It; this result Is More
pronounced for reactive pollu-
tants.
The effect of changing the Mix-
ing depth 1s not uniform over
the Modeling region; It varies
fron place to place.
The effect on ground-level con-
centrations of changing the
Mixing depth Is roughly the sane
as that of changing the wind
speed, as would be expected from
a dimensional analysis.
The effects of varying the
radiation Intensity are tine
dependent.
The effect of changing light
intensity Is as significant
as that of changing wind speed.
The study results are summar-
ized by the foilowlno ranking
of the relative iMportance
of the Input parameters (A •
Most important and 0 - feast
important):
-------
Table ,V-t2 (Continued)
Study
Model Version
and Attributes
Sensitivity Analysis
Variations
Influence on ttedel Predictions
Remarks
Reynolds et il. (1976)
SAI photochemical Model:
1973 version [see Liu
et •!. (1976i)]
Anderson et •!. (1977)
SAI photochemical mdel:
1977 version
31 step carbon bond
chemistry
3-0 Mind field
Lower microscale
layer
Lanb and Liu c!1ffu-
slvlty algorlttais
30 K 30 i 7 grid
SHASTA numerical
method
Surface rcMovat
Three-dimension*1
Initial conditions
Uniform wind velocities
with height were conpared
with vertical variation
in horizontal winds given
by a power law formulation
Hind speeds were reduced
by 33X
Nixing depths were re-
duced by 33X
Hind speeds and Mixing
depths were both reduced
by 331
Emissions in suburban
areas surrounding Denver
were reduced ZSX with
weighted emissions in-
crease) in other areas
to make overall regional
missions equivalent to
those in the bast case
Parameter or
Variable
Hind speed
Horizontal
dlffusivlty
Vertical
dlffusivlty
CO
A
D
C
NO
A
0
C
°3
A
D
C
NO,
A
D
C
The max (mum average percent-
age deviations were: 28.5S
for NO. 15X for NO?, 24S for
CO. and 14X for 03
The MxiNM average deviations
in pphu were: -0.35 for NO.
-1.1 for NO,. -4 for CO. and
-2 for 03 z
The MaxiMM deviations In pph»
were: 7.5 for NO, 15 for NO..
30 for CO. and 26 for 0} *
Maximum predicted ozone In-
creased bv 41; maximum area
for which [03] i O.OB ppm
Increased by 12S
predicted ozone In-
creased by 161; maximum
area for which [Oj] i O.OBX ppm
increased by 71
Naxlmm predicted ozone in-
creased by 331; maximum area
for which [0,] > 0.08 PPM
increased byJ30I
No difference occurred in the
time, location, or magnitude
of Maximum ozone concentration;
differences among predicted
ozone concentrations In all
runs were not More than 0.010
PPM in at Most one or two grid
cells
Mixing depth B • B B B
D A A B
Radiation
intensity
Emissions
rate
B
B B
The effects of including wind
shear were similar to those
of increasing surface wind
velocities by roughly 2SS
because velocities within the
mixed layer are Increased be-
tween 20 and 70S of the surface
values as a result of shear
A synergiSM exists between
wind speed and Mixing depth
In each scenario, no more than
7S of the region-wide emissions
were redistributed; changes of
this size in the spatial dis-
tribution of emissions has
little effect on secondary pol-
lutants such as ozone
-------
Table V-2 (Continued)
Study Group
Klllus (197?)
(private coMMjnicatlon)
Model Version
•IK! Attributes
SAI photochemical model:
1977 version (see
Anderson et •!. (1977)
for node! attributes]
Sensitivity Analyses
Variations
{Missions in the Denver
Metropolitan area were
reduced 17.51 with a pro-
portional increase in
suburban areas to Mke
regional tMisslon* levels
equivalent to those In
the base case
Grid spatial resolution
MBS relaxed from 1 x 1 mile
to 2 x 2 Miles
Influence on Model Predictions
No difference occured in the
liMe. location, or Magnitude
of MaxtMM ozone concentration;
differences aMong predicted
ozone concentrations in all
runs were not More than 0.010
PPM in at Most one or two grid
cells
The coarser grid resolution led
to no noticeable change in the
tine to peak NO. NOj. and Oj
concentrations; the Magnitude
of peak concentrations MBS re-
duced for NO (691). N02 (21t),
and 03 (131)
irks
By the tiMe ozone fonts. Its
precursors have been distri-
buted over a Much greater area
than their source regions;
accordingly, the Influence of
Increased grid size on ozone
predictions should'be less than
that for priMary pollutants
such as NO
Anderson (1977)
(private cnneunication)
SAI photochemical rndel:
1977 version [see
Anderson et al. (1977)
for Model attributes]
NO emissions fro* a point
source Mere Increased by
20f (note that the source
contributes roughly U of
the Denver regional NO.
burden)
The MaxiMUM Impact of increased
source missions anywhere in
the Modeling region was an in-
crease in hourly averaged NO
and NOj concentrations (12 and
5 ppb, respectively) and a
decrease In 03 (-4 ppb)
The effect was decidedly local
and did not Influence peak
oxldant concentrations
-------
CHAPTER VI. MODEL APPLICATION FOR AIR QUALITY PLANNING
Model application refers here to the use of an air quality model to
evaluate a specific air quality problem and to formulate control strategies
which, when Implemented, will ameliorate 1f not eliminate the problem. From
a regulatory viewpoint, the question the model application phase attempts to
address 1s: how much control 1s needed to reduce hydrocarbon emissions to a
level at which the ozone standard 1s attained? An ancillary question is:
what effect does nitrogen oxide control have on ozone levels? The air quality
planner would also Hke to know, for a given overall regional emission reduc-
tion, what source categories are most effective to control and to what level.
All of these questions can be addressed in an Airshed Model Application.
Before formulating control strategies, the air quality planner will
want to know what the situation 1s likely to be in future years, 1n terms of
ozone levels, if no additional controls are adopted. This 1s determined by
applying the Airshed Model using the baseline projection emission inventories
described in Chapter III. Data Needs. A preliminary estimate of the level
of additional control needed can then be made.
Two approaches have been used to formulate candidate control strategies,
one more rigorous than the other. The simpler approach is to use the
i
Airshed Model in a so-called "sensitivity" mode whereby Incremental across-
the-board emission reductions are made and the resulting ozone concentrations
are simulated. By employing this procedure, one attempts to "narrow in" on
the overall emission reduction needed and to subsequently select the control
measures necessary to provide that reduction. The preferred procedure is to
92
-------
simulate actual control measures in the context of an overall strategy.
By simulating alternate strategies, an optimum strategy may be Identified.
Whichever approach 1s taken, the final set of control measures should be tested
with the Airshed Model 1n order to demonstrate attainment.
A. PREPARATION OF FUTURE YEAR EMISSIONS
Whether a simulation of future air quality without additional controls
or whether a simulation of candidate control strategies 1s to be conducted,
one must first project the corresponding level of emissions resulting from
growth, controls, or both. In the former Instance, the projected emission
Inventory 1s called a baseline projection while 1n the latter 1t 1s called
a strategy projection. Normally one baseline projection 1s performed for
each projection year, usually 1987, and a later year (1995 for example).
However, some urban areas may want to examine the effects of alternative
levels of economic and population growth. In such cases, multiple baseline
projection scenarios may be needed.
Several strategy projections will also need to be performed for each
projection year. These projections reflect varying combinations of
Individual control measures, such as RACT (reasonably available control
technology), LAER (lowest achievable emission rate), I/M (Inspection and
maintenance), TCM's (transportation control measures), and possibly others.
The strategy projections also reflect which sources would be controlled
and when the controls would come Into effect. If only final strategy
projections are to be performed, having been formulated through use of the
Airshed Model 1n a "sensitivity" mode, only one or two projections may need
i
to be made. On the other hand, 1f a final control strategy 1s formulated
by simulating specific candidate strategies, a considerable number of
alternate strategy projections will need to be made.
It may also be desirable to examine a candidate strategy in the base year,
93
-------
As discussed 1n Chapter IV, Preparation of Model Input Data, the
EMISSIONS and PTSOURCE files contain hourly emissions of eight pollutants.
Furthermore, these emissions are 1n grldded form on the EMISSIONS file,
there being no distinction between one source and the next. Obviously,
neither baseline nor strategy projections can be performed at this level.
As discussed 1n Chapter III, Data Needs, stationary source projections
are made on an annual basis by source category. Mobile source projections
are made for the entire transportation network down to the link level 1n terms
of average dally traffic and peak hour traffic. The emissions, whether sta-
tionary or mobile, are recalculated and are disaggregated temporally. Hydro-
carbons and nitrogen oxides are apportioned to the five carbon bond categories
and to nitric oxide or nitrogen dioxide, respectively. Finally, the mobile
source, point source, and area source emissions are assigned or allocated to
the appropriate grid squares. Essentially then, performing a projection of
emissions for a future year 1s equivalent to regenerating the hourly grldded
Inventory for model Input.
B. FUTURE YEAR MODEL SIMULATIONS
Once the Airshed Model has been verified using one or more days of
aerometric data and once the baseline emission projections have been performed,
one Is ready to simulate future ozone levels. Preferably more than a single
day has been verified. If so, simulation of future year baseline emissions and
subsequent formulation of control strategies should usually be based on the
day having the highest ozone concentration. However, 1f the highest day
occurred not as a result of urban generated ozone but rather as a result of
transport, one will also want to formulate and evaluate control strategies on
a day for which urban generated ozone predominates.
94
-------
In addition to the projected emissions data, the starting points for
assembling model Inputs for future year simulations are the air quality
and meteorological data files for the day or days selected for control
strategy analysis. All modifications made to these files during the course
of the verification analysis should be reflected 1n the Input files used
for future year simulations. The meteorological files, namely WIND, DIFFBREAK,
REGIONTOP, TEMPERATUR, METSCALARS, and TERRAIN, are used directly and are not
further modified In any way. In contrast, the air quality files, namely
AIRQUALITY, BOUNDARY, and TOPCONC, must be adjusted to correspond to antld-
u
pated future year conditions, as discussed below.
1. Baseline Simulations
Simulating baseline conditions 1n a future year corresponds to a scenario
1n which growth occurs but no controls other than those having previously
been adopted are considered. Data preparation programs are run using the
projected baseline emissions to create the EMISSIONS and PTSOURCE files for
model Input. (These files were described previously 1n Chapter IV, Preparation
of Model Input Data). The files containing air quality data may or may
not require adjustment. Under a baseline scenario, sources outside the urban
area may have no controls beyond those 1n place during the base year. In this
case, an assumption could be made that transport at the upwind edge of the
modeling region and aloft remains the same as 1n the base year, thereby
requiring no adjustments to the BOUNDARY or TOPCONC files.
The AIRQUALITY file generally should be adjusted, however. This file con-
tains the Initial ambient concentrations for each pollutant at the beginning of
the simulation. Initial conditions could be assumed to be proportional to the
".}-.'•' °
urban regional emissions. Future year Initial conditions of hydrocarbons
could then be approximated by taking the ratio of the regional future year
95
-------
hydrocarbon emissions to the regional base year emissions and applying
this factor to the Initial conditions for all five carbon-bond categories.
However, natural background levels should be subtracted out first so that
only the anthropogenic portions of the Initial conditions are adjusted.
In addition, 1f any pronounced temporal variation exists 1n the emissions
ratio, this should be taken Into account. Nitric oxide and carbon monoxide
Initial conditions could be approximated similarly using NO and CO emissions,
^
respectively. Unfortunately* there 1s no such direct method for estimating
future ozone and nitrogen dioxide Initial conditions or those of other
secondary photochemical pollutant species (PAN and BZA). Since the accumu-
lation of these pollutants 1n the atmosphere 1s generally limited by the
amount of reactive hydrocarbons emitted, use of a future year to base year
hydrocarbon emission ratio could serve as a surrogate factor.
Model results may be presented as a printed map of ozone concentrations
as shown 1n Figure VI-1 or as ozone Isopleths as shown 1n Figure VI-2. The
Airshed Model Itself produces printed maps only; one number 1s plotted for
each grid cell. Maps may be generated for a selected hour at any vertical
grid cell level, but normally only ground-level concentrations are of Interest
when evaluating the need for controls. Computer plotting routines external
to the model are required 1n order to display ozone Isopleths.
2. Control Strategy Simulations
As Indicated earlier, two different approaches to control strategy develop-
ment may be taken. One 1s to use the Airshed Model 1n a "sensitivity" mode.
Across the board reductions 1n hydrocarbon and/or nitrogen oxide emissions
are made to varying levels. The principal advantage of this method 1s that 1t
Is relatively easy to do. The principal disadvantage 1s that 1t Ignores the
capability of the model to evaluate the effect of significant changes 1n the
96
-------
POLLUTANT: 03 DAY: 159 HOUR: 14-15
AVERAGE AMBIENT CONCENTRATIONS
-------
MIRTH
vo
00
•••••••••••••••••••••••••••••••tMi ••••••••••
30
SflUTH
FIGURE VI-2. EXAMPLE OF .A PLOTTED CONTOUR MAP OF GROUND-LEVEL OZONE CONCENTRATION
ESTIMATES. (Tesche and Burton, 1978).
-------
locations of emission sources and 1n the reactivity of the emissions. The
final strategy selected on this basis may either be more stringent than neces-
sary or less stringent depending on the manner In which the overall emission
reduction 1s actually distributed among sources, whether elevated or ground-
level, more or less reactive, upwind or downwind. The preferred method 1s
\
to project the effects of different combinations of control measures on future
year emissions, on a source by source basis as described above 1n Section
A, Preparation of Future Year Emissions. This method allows one to use the
Airshed Model to Its full potential.
Whichever approach 1s taken, 1t 1s helpful to have some notion before-
hand of the approximate level to which total regional hydrocarbon emissions
must be limited 1n order to meet the ozone standard. The Empirical Kinetic
Modeling Approach (EKMA) offers a simple technique (EPA, 1977c). Using the
EKMA standard curves, which show the relationship between ambient levels of
hydrocarbons and nitrogen oxides and the level of ozone under a set of standard
conditions, one may estimate an approximate percent hydrocarbon reduction
needed to reach the ozone standard. Two pieces of Information are required:
the ozone "design value" and the NMHC/NO ratio. The design value, as
n
described 1n "Guideline for the Interpretation of Ozone A1r Quality Standards"
(EPA, 1979), 1s the ozone concentration which has an expected exceedance of
once per year. The design value 1s determined statistically Irrespective of
the particular day selected for control strategy analysis. However, the NMHC/NO
ratio should correspond to the mean of the urban (as opposed to the upwind
or downwind) air quality monitoring sites for the 6 to 9 a.m. period on the
day or days selected for control strategy analysis. Once an approximate
percent hydrocarbon emission reduction for the base year 1s estimated using
the EKMA curves, this may be translated Into a percent reduction for any
99
-------
particular baseline projection year. This 1s done by first computing the
reduced base year regional hydrocarbon emission total and dividing by the
projection year baseline emission total. This of course 1s subtracted from
1.00 and then multiplied by 100 to obtain the percentage reduction. This
provides a first estimate of the overall level of control that may be needed.
With this Information one can begin to Identify the control measures that may
be necessary and to formulate candidate control strategies.
Boundary conditions for the modeling region are Important regardless of
the approach taken. Equity requires that a downwind city should not have to
exercise additional controls merely because an upwind city falls to meet Its
responsibilities to attain the ozone standard. Therefore, the air quality
planner will want to assume the upwind city 1s meeting the standard. How
then can the boundary conditions be made to reflect this? An Ideal approach
would be to use a regional photochemical model. A regional model 1s now
under development by the Environmental Sciences Research Laboratory, U. S.
Environmental Protection Agency, Research Triangle Park, North Carolina, but
1s not yet available. A simple technique 1s to assume that the change in
ozone transport aloft and along the upwind edge of the modeling region 1s
proportional to the change 1n ozone levels 1n the upwind city. Stated
another way, the ratt& of the upwind city "design value" for ozone during
the base year to the ozone standard 1s the same as the ratio of the ozone
boundary value 1n the base year to the boundary value 1n a future year. If
the maximum ozone level for the upwind city 1s known for the specific day
being modeled, that value should be used in place of the design value.
Transported precursors (hydrocarbons and nitrogen oxides) should probably not
be reduced unless it can be demonstrated the upwind city 1s the source of the
emissions. Generally, the relatively low levels of transported precursors
100
-------
normally observed are thought to have their origin 1n nonurban areas
Immediately upwind of a metropolitan area. Once an overall reduction 1n
transported ozone has been estimated, the BOUNDARY and TOPCONC files must
be recreated to reflect the change.
If the sensitivity approach 1s used to develop control strategies, the
emission reductions selected should generally correspond 1n an overall sense
to specific control measures. For example, preliminary analyses may Indicate
hydrocarbon reductions 1n the 25 to 50 percent range, relative to future year
baseline emissions are needed. Rough estimates of the effect of "RACT" on
point sources might show that a 22 percent reduction 1n total regional hydro-
carbons could be achieved while "LAER" might provide another five percent.
Application of a strong "I&M" program might reduce regional emissions by four
percent and a combination of "TCMs" another two percent. In this example,
one might select overall reductions of 22 percent (RACT) 33 percent (LAER/
I&M/TCMs) and 50 percent. The gap between 30 and 50 might, for Instance, be
ascribed to new source review Incorporating a stringent offset policy.
Across-the-board emission reductions such as these are readily accomplished
through the use of techniques provided 1n the data preparation programs. The
EMISSION FACTORS packet 1s suitable for this purpose. Initial concentrations
contained on the AIRQUALITY file should also be reduced by the same factor as
the emission rates. This is done by adjusting the station readings used as
input to the data preparation program; the program is then reexecuted to
create a new AIRQUALITY file.
Assuming the Airshed Model simulation runs show that the ozone standard is
met and exceeded (i.e., "bracketed") within the range of hydrocarbon reductions
selected, one may Interpolate between runs to estimate the overall percent
reduction required to Just reach the standard. It 1s then necessary for the
air quality planner to select what combination of control measures applied to
101
-------
which sources with what level of control are needed to attain the standard.
Alternate sets of measures should be tentatively selected because the key
step 1n the sensitivity approach 1s confirming one of these strategies by
means of actual Airshed Model simulations. This 1s done by performing detailed
emission projections (as described above 1n Section A. Preparation of Future
Year Emissions) followed by running of the Airshed Model. The model should
be run for each of the high ozone days previously verified. Hopefully, one
of the strategies chosen will show attainment on all test days. If not, more
stringent measures are needed and the process 1s repeated.
i
The alternate and preferred approach differs primarily In the manner
by which control strategies are formulated. Rather than making area-wide
emission reductions, one performs detailed projections of candidate strategies
composed of various control measures individually or in combination. Com-
bining them 1n different proportions to different source types, one may arrive
at any number of candidate strategies. For example, "RACT" reductions might
be applied to refineries and surface coating operations but not to service
stations in one strategy while another strategy might specify "LAER" reduc-
tions for all three. These candidate strategies are then simulated using
the Airshed Model. The results are then evaluated 1n the context of the
strategy's potential for attaining the ozone standard.
Once the candidate strategies have been simulated and the results
analyzed, the air quality planner is in a good position to select those
control measures and the source types to which each should apply which 1n
combination will attain the standard. If the candidate strategies have been
carefully selected in relationship to one another, a final strategy may
be arrived at readily with a minimum of trial and error while assuring,
when Rested, attainment will indeed be shown.
102
-------
Regardless of the approach taken, once a "final" strategy has been
selected, 1t should be tested to confirm that 1t will provide for attain-
ment of the ozone standard. A detailed projection of the effect of the
strategy on emissions as previously described should be performed. Air-
shed Model simulations are then conducted on several previously verified
high ozone days, not just the one or two days adopted- for strategy formu-
lation, 1n order to test the strategy under several meteorological and air
quality scenarios. This 1s Important because differing levels of ozone
transport from upwind urban areas and differing levels of ozone carry over
from a previous day, as well as differing meteorological conditions during
the days modeled, will all Influence the effectiveness of the particular control
strategy selected.
Airshed Model results from control strategy simulations may be presented
1n various ways. In addition to the printed maps and isopleth plots men-
tioned previously, graphical techniques for comparing strategies directly
are useful. Figure VI-3 shows a "deficit enhancement" plot on which Isopleths
of the difference 1n ozone concentrations between two control strategies or
between a control strategy and a baseline projection are shown. Having such
a plot, the magnitude and spatial distribution of the relative effect of a
particular control strategy may be seen at a glance. Figure VI-4 Illustrates
the temporal effects of a sensitivity run on maximum ozone levels over the
course of simulations on two different days. In this example, a 30 percent
reduction 1n both NOX and hydrocarbon emissions reduced the peak ozone con-
centration about 15 percent. However, the timing of the peak concentration
1s unchanged.
The ozone Isopleths resulting from different control strategies may be
analyzed further by determining at hourly Intervals the area over which
103
-------
NBRTH
20
30
SBUTH
FIGURE V1-3. DEFICIT ENHANCEMENT PLOT ILLUSTRATING THE PREDICTED EFFECT OF A CONTROL STRATEGY
(Tesche and Burton, 1978).
-------
i"
*
s
5 is
I.
II I
O—-O »ASE CASE
O——a 30 KRCENT KOUCTION
IN ALL EMISSIONS
3 AUGUST 1976 METEOROLOGY
I I I I
I"
10
i
0
28 JULY 1976 METEOROLOGY
! f
B
1 ! i
\
5
5
Ttat of toy of NMrly lotwttl
FIGURE VI-4. TEMPORAL EFFECTIVENESS OF A SENSITIVITY RUN ON PEAK OZONE CONCENTRATIONS
(Anderson, et al., 1977).
-------
estimated ozone levels exceed various threshold values. The results may then
be summarized graphically as shown 1n Figure VI-5. Differences between two
otherwise similar strategies may be readily shown 1n this way. For example,
one strategy might achieve greater reductions over a greater area than another
.>
even though both provide nearly the same reduction 1n peak ozone levels.
(Note that 1n this example, controls on NO emissions from power plants cause
A
higher urban ozone levels, a phenomenon which follows from the scavenging effect
nitric oxide has on ozone.)
The results from Airshed Model simulations may also be translated Into
population exposures as shown on Figure VI-6. The model generates hourly
ozone concentrations 1n each ground-level grid cell. These can be used In
conjunction with grldded population values (previously estimated during the
development of area source grid allocation factors) to compute the number of
person-hours above various ozone threshold values. The population exposure
curves are then drawn by plotting these values for each strategy or each year
against the corresponding ozone concentration.
C. INTERPRETATION OF MODEL RESULTS
Two Issues arise in regard to the application of the Airshed Model and
other photochemical models to control strategy analysis. The first Issue
relates to model performance. No model can be expected to give results which
are totally free of bias. Even on days for which a successful verification
has been achieved, some residual bias will undoubtedly remain. The second
Issue 1s the comparability of the model results to the ozone standard. The
standard 1s expressed 1n terms of an expected exceedance while the model
results represent concentrations for a particular high ozone day on which
the model has been run. Both of these Issues should be addressed when
106
-------
§
2
•M
§
U
§
o
01
i
O tf)
•M 10
15
10
5
0
40
^^ ^0 *
2-3 .
o
T-
*
o
°=* 20
01 -o
10
0)
3
pphm
10 11
TT 17
12
TI
13
T?
14
15
16
T7
Time of Day (PST)
pphm
BASE CASE
POWER PLANT
MOBILE SOURCE
pphm
FIGURE VI-5. SIMULATED AREAL INFLUENCE OF TWO EMISSION CONTROL STRATEGIES
ON 03 CONCENTRATIONS ABOVE THREE CONCENTRATION LEVELS
(Tesche and Burton, 1978).
107
-------
o
00
I
W
I
1976
1985
8 10 12 14 16 18 20 22 24 26 28
Oiont Conctntratlon (ppiw)
FIGURE VI-6. EXAMPLE OF THE RESULTS OF A POPULATION EXPOSURE ANALYSIS USING THE AIRSHED MODEL
(Anderson, et al., 1977).
-------
applying the model to control strategy analysis to avoid either over-
stating or understating the level of control required to meet the ozone
standard.
The best way to eliminate model bias 1s to adjust the model Itself
to remove the bias at Us source. However, such "fine tuning" of the model
could be quite time consuming and may be Impractical under the time and
resource constraints Imposed upon the modeling effort. Calibration offers
an alternative to reformulating the model. Calibration attempts to adjust
the model ozone estimates themselves to better correspond with ozone
observations. Calibration does not necessarily Improve the accuracy of the
model; rather 1t simply ttrles to compensate for any bias present 1n the
results. Calibration 1s not a substitute for good model performance.
The second model Interpretation Issue arises when the prototype day has
a different maximum ozone concentration that does the design day. The
prototype day 1s the particular day on which the model 1s run. The design
day 1s a hypothetical day on which the ozone design value occurs. The design
value 1s the ozone concentration which has an expected exceedance of once
per year. Methods for determining the design value are discussed 1n
"Guideline for the Interpretation of Ozone A1r Quality Standards" (EPA,
1979).
There are no assurances that a day with an ozone concentration equal
to the design value will occur during the three-month aerometric data
collection study. Typically, the day or days selected for modeling will
have a maximum observed ozone concentration less than the design value.
Therefore, the ability to demonstrate on a prototype day that the ozone
standard 1s not exceeded under a given emission control strategy does
not necessarily assure that the ozone standard will not be exceeded on
109
-------
the design day. The only entirely satisfactory way to address the design
day Issue 1s to actually evaluate a given strategy using the Airshed Model
on the design day. However, given that the prototype day does not corres-
pond to the design day, an objective technique 1s needed to relate model
estimates on the prototype day to the design day. The question becomes, how
can the model estimates be adjusted so they can be used for comparison with
the ozone standard? An attempt at answering this question has been made by
the Association of Bay Area Governments (ABAG, 1979).
110
-------
CHAPTER VII. RESOURCE REQUIREMENTS FOR
AN AIRSHED MODEL STUDY
Use of the Airshed Model 1n an urban area for control strategy planning
1s an expensive undertaking. As described 1n earlier chapters, a large data
base needs to be amassed first. Emissions data must be collected, compiled,
reviewed, and analyzed. Field studies to gather the necessary meteorological
and air quality data must be planned and executed. The resulting data must
be compiled, reviewed, validated, and analyzed. Each of these two data col-
lection activities Involve long lead times and are generally done largely
i
under contract. Modeling Itself requires specialized skills and experience
not commonly found 1n local pollution control agencies and 1s therefore
normally a contract activity. Modeling verification may be a lengthy process
but one which must be accomplished before control strategy evaluation and
testing can begin. Execution of the Airshed Model requires the use of a
large high speed computing facility. Running times are long and computer
costs are substantial.
A. MODELING PERSONNEL
While the number of persons needed to perform the modeling verification
and control strategy analysis phases of the overall study is not necessarily
large, a diversity of knowledge and experience 1s required. The various
disciplines needed Include:
- engineering
• computer science
- meteorology
- chemistry
- statistics
111
-------
Because personnel normally follow disciplinary lines, the modeling team
should Include at a minimum an air pollution engineer, a systems analyst
or programmer, an air pollution meteorologist, an atmospheric chemist,
and a statistician. The engineer and computer specialist are often Involved
1n most aspects of the modeling. The extent of involvement of the meteor-
ologist, chemist, and statistician depends on their individual background
and experience 1n modeling. Their abilities are essential 1n certain key
phases of the study, yet their special expertise is not necessarily required
throughout the entire study. For example, the skills of the meteorologist,
chemist, and statistician are not normally needed during the control strategy
evaluation phase. However, any one of the team participants, regardless of
their particular disciplines, may take on the lead role in the study depend-
ing on his or her experience in photochemical modeling.
The air pollution engineer should have experience in applying models and
should be familiar with emission inventory practices and techniques. During
the control strategy evaluation phase, he has primary responsibility for pre-
paring strategy emission projections, for preparing the model Input data, and
for evaluating the model results. During the model verification phase, he 1s
involved 1n evaluating verification problems and conducting sensitivity
analyses, particularly as related to emissions data.
The air pollution meteorologist should have a background 1n boundary
layer meteorology and modeling. It 1s his responsibility to prepare the
meteorological data for model Input. This Includes specifying the wind field,
mixing heights, and stability data for the modeling region. He 1s heavily
Involved in the verification phase in evaluating the effect of meteorological
variables on model performance. An 1n-depth analysis of the meteorological
112
-------
data base and the use of several data preparation schemes may be needed
before satisfactory results are achieved.
The atmospheric chemist should have a background 1n ambient monitoring
and the chemistry of ozone formation. He should be especially familiar
with the carbon-bond mechanism and how 1t treats various organic compounds.
The atmospheric chemist 1s responsible generally for assisting 1n the
assessment of the reasonableness of the emission Inventory and specifically
for reviewing how the organic emissions from various source types have been
assigned to the five carbon-bond categories. He 1s also responsible for
preparing the ambient nonmethane hydrocarbon data. The chemist plays a key
role 1n evaluating air quality data aloft and upwind 1n order to prepare
model boundary conditions. He also assists 1n the preparation of Initial
conditions. The atmospheric chemist Joins 1n the verification effort to
determine whether the air quality data 1s 1n need of re-examination and
whether the model chemical kinetics are performing properly.
The computer specialist 1s the backbone of the modeling team. It 1s
his Job to carry out many of the data manipulations necessary to create the
files requested by the other project team members. Frequently, this may
Involve modifications to the existing data preparation programs or entire
new programs may need to be written. Most of the data manipulations neces-
sary for performing strategy emission projections, followed by the necessary
splitting, grlddlng, and recomblnlng, must be done using computer processing.
Obviously, the computer specialist must be thoroughly familiar with the
computer system on which the model 1s to be exercised. He must be know-
ledgeable concerning system design, data storage, tape handling procedures,
and structured programming. The computer specialist must, of course, know
113
-------
how to operate the large Airshed Model system of programs and files.
Finally, he should have a working knowledge of computer plotting soft-
ware so that model outputs (and Inputs) can be examined and evaluated
graphically.
The statistician 1s primarily Involved in evaluating model perform-
ance. It 1s he who 1s responsible for Identifying the performance measures
to be used, for applying, and for Interpreting such measures. He assists
1n examining the model results to isolate probable causes of verification
problems. The statistician works with the computer specialist in manipu-
lating the data, either numerically or graphically. He may also be called
upon at the beginning of the modeling study to assist in examining and
analyzing the data base, particularly the aerometric data.
B. COMPUTER FACILITY
The Airshed Model computer program 1s written entirely 1n FORTRAN.
i
The executable module requires 61K 36-BIT words of memory for operation on
a UNIVAC 1144. On the same computer, running time for the model 1s approxi-
mately 15 minutes per hour of simulation for a modeling region of 300 grid
squares and four vertical layers, I.e., 1200 grid cells altogether. Since
running time depends primarily on the size of the modeling region, more
grid cells would require proportionally more time. However, this time can
also vary somewhat due to the shorter time steps necessary for Increasing
wind speeds. If the user-defined modeling region is too large to fit into
the available core of the host computer, a segmentation process 1s avail-
able which divides the region Into one or more rectangular segments. Using
segmentation, the data arrays for all segments are maintained in secondary
storage while the arrays for the segment being processed reside 1n core.
114
-------
The times and core requirements given above reflect a modeling region
divided Into two segments.
Initially the model requires the operation of eleven preprocessor
programs. Each of these programs performs a similar function 1n that each
prepares and formats the various types of data required for Input to the
model. Each program requires approximately 60K 36-BIT words of memory for
operation on a UNIVAC 1144. Also, each program requires an average of
approximately two minutes of running time. The files produced by these
programs can be written on either magnetic tape or disc. Eleven output files
are produced for Input to the model. In addition to these eleven, the seg-
mented modeling region described above requires the use of seventeen external
mass-storage files requiring an overall total of approximately 725,000
36-BIT words.
C. PROJECT TIMETABLE
A project the size of a photochemical modeling study using the Airshed
Model poses considerable uncertainty not only 1n regard to Its cost, but also
in regard to the time required for Its completion. Figure VII-1 represents
the kind of study timetable one may expect. Overall, a study using the
Airshed Model may be expected to take three years or more.
Once the local pollution control agency has made the decision to perform
an Airshed modeling study, Initial planning begins, this may take about
six months. An analysis of the existing emissions data base and the
historical meteorological and air quality data must be carried out. With
this Information, decisions can be made on the dimensions of the modeling
region and on the exact data needs for the study. The adequacy of the
existing monitoring network and the need for additional meteorological and
air quality monitoring sites and Instrumentation must be assessed. The
115
-------
YEARS 0 1 2
i i
MONTHS
TASKS '
INITIAL PLANNING
DATA COLLECTION
Contract Planning
AQ/MET Field Study
AQ/MET Data Compilation
Emission Inventory Data Collection
Emission Inventory Compilation
MODELING
•> Contract Planning
Performance Evaluation
Strategy Evaluation
1 1 1 1 1 1 1 1
1 3 6 9 12 15 18 21 24
i
*
i i
i i
•
I1
3
27 30 33 36
o , ,,.n|
I h.
Figure VI1-1. A Photochemical Modeling Study timetable.
-------
requirements for upper air meteorological and air quality data must be estab-
lished. The extent to which the existing stationary source emission Inventory
meets the needs of the study must be assessed and additional data needs must
be Identified. The level of effort required to take the existing transportation
data base and use 1t for generating the mobile source emission Inventory must
be estimated. The assistance of a contractor familiar with the Airshed Model
and Us requirements 1s a valuable part of the Initial planning effort. Toward
the end of the Initial planning phase, 1t should be possible to make reasonable
estimates of the resources required to perform the study.
Soon after the Initial planning, the local agency will want to decide on
the extent of contractor Involvement. This depends on the In-house man-hours
available to devote to the project and the availability of outside funding.
Contractors are oftentimes heavily Involved 1n the data collection effort.
Depending on agency contracting procedures, contract solicitation, selection,
and award can require seven months of elapsed time, if all goes well.
Once one or more contractors have been selected to operate supplementary
surface monitoring sites, collect aircraft and upper air meteorological data,
and analyze hydrocarbon grab samples, the field study can begin. For a ,
three month data-collection period, one should probably count on at least
two months of preparation time. New monitoring stations and associated
Instrumentation must be acquired, Installed, and readied. Aircraft must be
equipped, the flight crew assembled, and flight plans drawn up. Radiosondes,
plbals, and associated tracking equipment must be obtained and release sites
prepared. Existing surface monitoring stations may need to be reequlpped
or overhauled. With the onset of the "oxidant season," data collection begins.
After it 1s collected, the data must be validated, archived, reviewed, and
analyzed. Before the data is ready for modeling, another seven months will
likely have passed.
117
-------
Concurrent with the aerometrlc data collection effort 1s the emissions
data collection effort. Agency stationary source files must be reviewed
and updated as necessary. Area source emissions must be determined for each
county 1n the modeling region and allocation factors must be developed to
allocate them to the grid system. Temporal profiles must be developed for
each source as do profiles for apportioning hydrocarbon and nitrogen oxide
emissions to the five carbon-bond categories and to NO and N02- Diurnal traffic
distribution data must be developed, VMT estimates made, and speeds computed
for the transportation Inventory. Vehicle operating characteristics must be
estimated. Approximately six months may be required for collecting these
various data. Another six months may be needed to ready network emissions
models, to prepare other computer software, and then to generate the hourly
gridded emission Inventories for model Input.
Well before the emissions and aerometrlc data bases are expected to be
ready for modeling, the agency should obtain the services of a modeling con-
tractor. As for the data collection effort, seven months may be required to
consummate a contract.
Model performance evaluation begins with an analysis of the data base.
The meteorological and air quality data must then be prepared for model Input.
Once the model 1s run, the results must be evaluated and any problems cor-
rected. This requires about nine months. The formulation of candidate
control strategies and the selection and testing of a final strategy are
open-ended activities which may last as long or longer than the model per-
formance evaluation.
D. OVERALL PROJECT COSTS
Table VII-1 shows a range of overall project costs, exclusive of com-
puter costs, broken down by major activities. All cost Hems are assumed to be
118
-------
Table VII-1. Photochemical Modeling Study Costs
Activity
PLANNING
DATA COLLECTION
Surface AQ/MET Data
Upper A1r MET Data
Aircraft AQ Data
Organic Species AQ Data
Emissions Data
Subtotal
MODELING
Performance Evaluation
Strategy Evaluation
Subtotal
TOTAL
Cost1
Low ($1000)
20
250
30
20
150
450
Computer Time*
100
80
180
650
H19h
40
550
100
150
40
850
1690
180
no
290
2020
Low (Hours) ^*
_
1 3
1 1
1
-
15_ 30_
17 35
7 12
JL li
1_5 28
32 63
1
Contract costs, excluding computer costs, 1979 dollars.
Tor a CDC 7600 machine.
119
-------
done outside the local agencies, under contract. Costs for managing the project,
which may require one or two man-years 1n and of Itself, are assumed to be
absorbed by the agency's own manpower resources and not require external fund-
Ing. Any tasks performed by the agency In-house will reduce project costs pro-
portionally. Also shown 1n Table VII-1 are approximate computer time requirements
for a machine such as the CDC 7600. Based on these estimates, total noncomputer
costs for a photochemical modeling study may range from a low of 600 thousand
dollars to a high of 1.9 million. Between 75 and 85 percent of this 1s for
data collection while the remainder 1s for modeling and for contractor assistance
in the initial planning of the project. In addition, at a unit cost for com-
puter time of two thousand dollars per hour on a machine such as the CDC 7600,
total computer costs may range from 65 to 125 thousand dollars.
Previous experience in planning photochemical modeling and aerometric and
emissions data collection studies shows that Initial planning costs may range
from 20 to 40 thousand dollars. This 1s the cost at the start of the project
for contractor assistance 1n planning the overall study and of determining
what the particular data needs are.
Collection of surface air quality and meteorological data 1s usually the
<*
largest single cost Item in a photochemical modeling study. As indicated in
Table VII-1, costs may vary from 250 to 550 thousand dollars or more. The
cost depends on the number of new monitoring sites required and on any
additional monitoring equipment needed at existing state and local agency
sites. The cost figures shown here assume that all new sites are Installed
and operated under contract. No costs are Included for operating existing
monitors since these costs are already being Incurred by state and local
agencies. In Philadelphia, (PA-NJ-DL-AQCR) for example, where an Airshed
Model application 1s underway, operation of five new sites for three months
cost 250 thousand dollars. This Includes data collection, compilation,
120
-------
validation, review, and analysis but does not Include the cost of Instruments,
which were provided to the contractor. These five sites were supplemented
by 17 state and local sites. Some computer time 1s also required to compile
and analyze the entire surface aerometHc data base.
Other cities where photochemical modeling studies are being planned are
New York, Boston, Baltimore, and Washington, D.C. Estimated costs for
surface aerometrlc data range from 550 thousand dollars 1n New York to 350
thousand 1n Baltimore. In Baltimore, four new sites are proposed to sup-
plement seven existing sites which are being upgraded while in New York
(NY-NJ-CN AQCR) the number of new and existing sites are 10 and 20 respec-
tively.
Upper air meteorological data consists of rawinsonde, plbal, and
acoustic sounder measurements. Among cities now being studied, cost estimates
range from 30 to 100 thousand dollars. Equipment and operating costs are
included. The high estimate of 100 thousand dollars is for Boston where no
sites now exist and a radiosonde and two plbal sites are planned. At the low
end, New York, a radiosonde site 1s available; two additional plbal sites are
planned. In Philadelphia, 50 thousand dollars was expended for a single
radiosonde site at which hourly plbal releases were also made.
Measurements of air quality data aloft using Instrumented aircraft
may or may not be taken as discussed in Chapter III. Data Needs. It has
been estimated to cost 150 thousand dollars for 15 in-flight days. This
also includes the cost of standby flight crews ready to fly during a typical
three month summer study. However, these are only operating costs and do not
reflect the cost of equipping an airplane with instruments.
121
-------
A cost of 30 thousand dollars has been estimated for collecting 200
grab samples and conducting a complete organic species analysis using gas
chromotography. The cost for a given study area depends on the number of
samples analyzed.
The cost of developing an hourly gridded emissions Inventory for model
Input 1s not so much dependent on city size, either 1n terms of geographic
area or population, but rather on the condition of the existing data base.
Contract costs may range from a low of 150 thousand dollars to a high of 850
thousand dollars for an Inventory of mobile and stationary sources. In
Baltimore, where the stationary source Inventory 1s relatively complete and
current, an estimated 70 thousand dollars 1s needed to derive the necessary
spatial and temporal resolution, to develop VOC and NO pollutant profiles,
A
and to perform two baseline projections. Another 95 thousand dollars 1s
needed to generate the mobile source emission Inventory, making the total
Inventory costs for Baltimore approximately 170 thousand dollars. In compar-
ison, 1n the Metropolitan New York (NY-NJ-CN) AQCR, over half of an estimated
850 thousand dollar total cost Is needed just to collect the data necessary
to develop a basic annual Inventory of point and area sources. Only 125
thousand of the total 1s needed for mobile sources. In the metropolitan
Washington area, a total cost of 250 thousand 1s estimated of which 140
thousand 1s for mobile sources and 110 for stationary sources. These costs
all Include the cost of computer processing. In Baltimore, 28 thousand
dollars has been budgeted for computer processing to generate mobile source
emissions for model Input from the transportation Inventory. In Tulsa, 19
thousand dollars was expended for stationary source emissions computer
processing.
122
-------
In addition to the basic work of collecting and assembling the emis-
sions data and manipulating 1t for model Input, 1t 1s necessary to oversee
closely the Inventory effort and to thoroughly review the emissions data.
This quality assurance function, 1f done under contract, may cost between
25 and 50 thousand dollars. However, these costs are not Included 1n
Table VII-1 where 1t has been assumed the local agency will perform this
function without external funding.
The cost for model verification and control strategies depends on the
number of Airshed Model runs and the size of the modeling region (and there-
fore the size of the data base). The same 1s true of the estimated computer
time. Modeling costs range from 180 to 290 thousand dollars. Model verifi-
cation Includes the cost of analyzing the data and preparing 1t for model Input
as well as the simulations themselves, exclusive of computer costs. Also shown
1n Table VII-1 are estimated computer time requirements. At least two trial
simulations are assumed necessary to achieve satisfactory model performance
on any given day; however, more may be required. Model verification on three
or four different days 1s also assumed. Control strategy analysis Includes
the cost of performing strategy emission projections and of running baseline
simulations and candidate control strategy simulations. Anywhere from five
to 15 runs may be required depending on what control measures are desired to
be evaluated. In addition, final strategy confirmation runs on three or
four days are also Included. It should be realized that 1n running a model
as large as the Airshed Model, a significant amount of computer time can be
expended on wasted runs, runs for which no meaningful output results, due to
various pitfalls and errors. The experience of one agency with a different
model was that for every successful run, an average of 2.5 runs had to be
made altogether (ABAC, 1979).
123
-------
REFERENCES
ABAC (1979). "Application of Photochemical Models 1n the Development of
State Implementation Plans, Volume I: The Use of Photochemical Models
1n Urban Oxldant Studies," EPA-450/4-79-025, Association of Bay Area
Governments, Berkeley, California.
Ames, J. et al. (1978). "The User's Manual for the SAI Airshed Model,"
draft report for Environmental Sciences Research Laboratory, Office of
Research and Development, U. S. Environmental Protection Agency, Contract
No. 68-02-2429, Systems Applications Inc., San Rafael, California.
Anderson, G. E., et al., (1977). "A1r Quality 1n the Denver Metropolitan
Region: 1974-2000," EPA-908/1-77-002 Systems Applications Inc.,
San Rafael, California.
BHggs, G. A. (1971). "Plume Rise: A Recent Critical Review," Nuclear
Safety. Vol. 12, pp. 15-24.
Bucon, H. W., J. F. Macko, and H. 0. Taback (1978). "Volatile Organic
Compound (VOC) Species Data Manual," EPA-450/3-78-119, KVB Engineering,
Inc., Tustln, California.
Demerjlan, K. L. (1976), "Photochemical A1r Quality Simulation Modeling:
Current Status and Future Prospects," Paper 16-1, International Con-
ference on Photochemical Oxldant Pollution and Its Control, U. S.
Environmental Protection Agency, Raleigh, North Carolina.
Duewer, W. H., M. C. McCracken, and J. J. Walton (1978). "The Llvermore
Regional A1r Quality Model: II. Verification and Sample Application
1n the San Francisco Bay Area," J. Appl. Meteor., Vol. 17, pp. 273-311.
Eaton, W. C., et al., (1979). "Study of the Nature of Ozone, Oxides of
Nitrogen, and Nonmethane Hydrocarbons 1n Tulsa, Oklahoma, Volumes I, II
and III," EPA-450/4-79-008a, b, c, Research Triangle Institute, Research
Triangle Park, North Carolina.
EPA (1979). "Guideline for the Interpretation of Ozone A1r Quality Standards,"
EPA-450/4-79-003, Office of A1r Quality Planning and Standards, Research
Triangle Park, North Carolina.
EPA (1978a). "User's Guide to MOBILE!: Mobile Source Emissions Model,"
EPA-400/9-78-007, Office of A1r, Noise and Radiation, U. S. Environ-
mental Protection Agency, Washington, D.C.
EPA (1978b). "Mobile Source Emission Factors, Final Document,"
EPA-400/9-78-005, Office of Transportation and Land Use Policy,
U. S. Environmental Protection Agency, Washington, D. C.
124
-------
EPA (1977a). "Compilation of A1r Pollutant Emission Factors, Third
Edition," AP-42 parts A and B, Office of Air Quality Planning and
Standards, U. S. Environmental Protection Agency, Research Triangle
Park, North Carolina.
EPA (1977b). "Procedures for the Preparation of Emission Inventories
for Volatile Compounds, Volume I," EPA-450/2-77-028, Office of A1r
Quality Planning and Standards, U.S. Environmental Protection Agency,
Research Triangle Park, North Carolina.
EPA (1977c). "Uses, Limitations and Technical Basis of Procedures for
Quantifying Relationships Between Photochemical Oxldants and Precursors,"
EPA-450/2-77-021a, Office of A1r Quality Planning and Standards, U. S.
Environmental Protection Agency, Research Triangle Park, North Carolina.
EPA (1976). "AEROS Manual of Codes," EPA-450/2-76-005t Office of A1r
Quality Planning and Standards, U. S. Environmental Protection Agency,
Research Triangle Park, North Carolina.
EPA (1975). "Comprehensive Data Handling System, Emissions Inventory/Permits
and Registration Subsystem (EIS/P&R) Program Documentation and Users
Guide/ EPA-450/3-74-045a as updated, Office of A1r Quality Planning
and Standards, Research Triangle Park, North Carolina
Engineering-Science (1979a). "Emission Inventories for Tulsa, Oklahoma
for SAI Model Application," draft report for Office of Air Quality
Planning and Standards, U.S. Environmental Protection Agency; Contract
No. 68-02-2584, Engineering-Science, McLean, Virginia.
Engineering-Science (1976b). "Emission Inventory Protocol Document [for]
Metropolitan Philadelphia Oxldant Study," draft report for Office of
A1r Quality Planning and Standards, U. S. Environmental Protection
Agency, Contract No. 68-02-2584, Engineering-Science, McLean, Virginia.
GHscom, R. W. (1978). "Regional A1r Pollution Study: Point and Area
Source Organic Emission Inventory," EPA-600/4-78-028, Rockwell
International A1r Monitoring Center, Creve Coeur, Missouri.
Hayes, S. R. (1979). "Performance Measures and Standards for A1r Quality
Simulation Models," EPA-450/4-79-032, Systems Applications Incorporated,
San Rafael, California.
HUlyer, M. J., S. D. Reynolds and P. M. Roth (1979), "Procedures for
Evaluating the Performance of A1r Quality Simulation Models,"
EPA-450/4-79-033, Systems Applications, Incorporated, San Rafael,
California.
Jerskey, T. N., and J. H. Seinfeld (1976). "Continued Research 1n Meso-
scale A1r Pollution Simulation Modeling—Volume IV: Examination of the
Feasibility of Modeling Photochemical Aerosol Dynamics," EPA-600/4-76-016d,
Systems Applications, Incorporated, San Rafael, California.
125
-------
Oerskey, T. N., et al., (1976). "Continued Research 1n Mesoscale A1r
Pollution Simulation Modeling—Vol. VII: Mathematical Modeling of
Urban Aerosol Dynamics," draft report for Environmental Sciences
Research Laboratory, Office of Research and Development, U. S.
Environmental Protection Agency, Contract No. 68-02-2216, Systems
Applications, Incorporated, San Rafael, California.
Kill us, J. P., et al., (1977). "Continued Research 1n Mesoscale A1r
Pollution Simulation Modeling—Vol. V: Refinements 1n Numerical
Analysis, Transport, Chemistry, and Pollutant Removal," draft report
for Environmental Sciences Research Laboratory, Office of Research and
Development, U. S. Environmental Protection Agency, Contract No. 68-02-2216,
Systems Applications, Incorporated, San Rafael, California.
Lamb, R. G. (1976). "Continued Research 1n Mesoscale A1r Pollution Simu-
lation Modeling—Volume III: Modeling of Mlcroscale Phenomena,"
EPA-600/4-76-016c, Systems Applications, Incorporated, San Rafael,
California.
Lamb, R. G., W. H. Chen, and J. H. Seinfeld (1975). "Numerico-Empirical
Analysis of Atmospheric Diffusion Theories," J. Atmos. Sc1.. Vol. 32,
pp. 1794-1807.
Lamb, R. G., et al., (1977). "Continued Research 1n Mesoscale A1r
Pollution Simulation Modeling—Vol. VI: Further Studies 1n the Model-
Ing of Mlcroscale Phenomena," draft report for Environmental Sciences
Research Laboratory, Office of Research and Development, U. S.
Environmental Protection Agency, Contract No. 68-02-2216, Systems
Applications, Incorporated, San Rafael, California.
L1u, M. K. and J. H. Seinfeld (1975). "On the Validity of Grid and
Trajectory Models of Urban A1r Pollution," Atmos. Environ.. Vol. 9,
pp. 555-574.
Liu, M. K., D. C. Whitney and P. M. Roth (1976). "Effects of Atmospheric
Parameters on the Concentration of Photochemical Pollutants," J. Appl.
Meteorol., Vol. 15, pp. 829-835.
L1u, M. K., et al., (1976a). "Continued Research 1n Mesoscale A1r
Pollution Simulation Modeling—Volume I: Analysis of Model Validity
and Sensitivity and Assessment of Prior Evaluation Studies,"
EPA-600/4-76-016a, Systems Applications, Incorporated, San Rafael,
California.
Ludwlg, F. L. and E. Shelar (1978). "Site Selection for Monitoring
Photochemical A1r Pollutants," EPA 450/3-78-013, SRI International,
Menlo Park, California.
MacCracken, M. C., et al., (1978). "The Llvermore Regional A1r Quality
Model: I, Concept and Development," J. App. Meteor.. Vol. 17,
pp. 254-272.
126
-------
MacCracken, M. C. and 6. D. Sauter, eds. (1975). "Development of an A1r
Pollution Model for the San Francisco Bay Area," "UCRL-51920 Vols.
1 and 2, Lawrence Livermore Laboratory, University of California,
Llvermore, California.
MacCracken, M. C. (1975). "User's Guide to the LIRAQ Model: An A1r
Pollution Model for the San Francisco Bay Area," UCRL-51983, Lawrence
Llvermore Laboratory, University of California, Llvermore, California.
MIlHgan, R. J., et al., (1979). "Review of NO Emission Factors for
Stationary Combustion Sources," EPA 450/4-79-021, Acurex Corporation,
Mountain View, California.
Pacific Environmental Services, Inc. (1979). "Procedures for the Prepara-
tion of Emission Inventories for Volatile Organic Compounds, Volume II
Emission Inventory Requirements for Photochemical A1r Quality Simu-
lation Models," EPA-450/4-79-018, Pacific Environmental Services,
Santa Monica, California.
Reynolds, S. D. (1977). "The Systems Applications, Incorporated Urban
Airshed Model: An Overview of Recent Developmental Work," Inter-
national Conference on Photochemical Oxldant Pollution and Its Control,
EPA-600/3-77-001b, Systems Applications, Incorporated, San Rafael,
California.
Reynolds, S. D., et al., (1979). "Photochemical Modeling of Transportation
Control Strategies. Volume I. Model Development, Performance
Evaluation, and Strategy Assessment," draft final report for Office
of Research, FHWA, U.S. Department of Transportation, EF79-37, Systems
Applications, Incorporated, San Rafael, California.
(1978). "Application of the SAI Airshed Model to the Evaluation
of Alternative Population Growth Forecasts for the South Coast A1r
Basin," EF78-124, Systems Applications, Incorporated, San Rafael,
California.
(1976). "Continued Research 1n Mesoscale A1r Pollution Simulation
Modeling: Volume II: Refinements 1n the Treatments of Chemistry,
Meteorology, and Numerical Integration Procedures," EPA-600/4-76-016b,
Systems Applications, Incorporated, San Rafael, California.
(1974). "Mathematical Modeling of Photochemical A1r Pollution—III,
Evaluation of the Model," Atmos. Environ.. Vol. 8, pp. 563-596.
(1973a). "Further Development and Validation of a Simulation
Model for Estimating Ground Level Concentrations of Photochemical
Pollutants," Systems Applications, Incorporated, San Rafael, California.
(1973b). "Mathematical Modeling of Photochemical A1r Pollution—I.
Formulation of the Model," Atmos. Environ.. Vol. 7, pp. 1033-1061.
127
-------
Reynolds, S. D.f T. W. Tesche, and L. E. Reid (1978). "An Introduction to
the SAI Airshed Model and Its Usage," draft report to Office of A1r
Quality Planning and Standards, U. S. Environmental Protection Agency,
EF78-53R3, Systems Applications, Inc., San Rafael, California.
Roth, P. M., et al.t (1974). "Mathematical Modeling of Photochemical
A1r Pollution--!!. A Model and Inventory of Pollutant Emissions,"
Atmos. Environ.. Vol. 8, pp. 97-130.
(1971). "Development of a Simulation Model for Estimating Ground
Cevel Concentrations of Photochemical Pollutants," APTD-0914 (Also
Appendices A-F, APTD-0908 thru 0913), Systems Applications, Incorporated,
San Rafael, California.
Schere, K. L. and K. L. Demerjlan (1978). "A Photochemical Box Model for
Urban A1r Quality Simulation," Proceedings of the Fourth Joint Con-
ference on Sensing of Environmental Pollutants, American Chemical
Society, 6-11 November 1977, New Orleans, Louisiana.
Tesche, T. W. and C. S. Burton (1978). "Simulated Impact of Alternative
Emission Control Strategies on Photochemical 0x1dants 1n Los Angeles,"
EF78-22R, Systems Applications, Inc., San Rafael, California.
THjonls, J. C. and K. W. Arledge (1976). "Utility of Reactivity Criteria
1n Organic Emission Control Strategies: Application to the Los Angeles
Atmosphere," EPA-600/3-76-091, TRW Environmental Services, Redondo
Beach, California.
128
-------
APPENDIX
TECHNICAL DESCRIPTION OF THE AIRSHED MODEL
-------
TABLE OF CONTENTS
A. The Basic Equation A-3
B. The Numerical Solution Procedure A-5
C. The Estimation of Turbulent D1ffus1v1t1es A-10
D. The Treatment of Atmospheric Chemistry A-15
E. The Treatment of Emissions A-25
F. The Treatment of Surface Uptake A-30
A-l
-------
TECHNICAL DESCRIPTION OF THE AIRSHED MODEL*
The development of the Airshed Model 1s described 1n numerous reports
which detail the derivation of the model and the steps within each algorithm
and subroutine. These reports were mentioned previously 1n the Foreword to
this document. Although quite useful for developing an understanding of the
model, their length makes their use difficult. This appendix provides a brief
technical description of the model so that users can become familiar with the
basic concepts and equations used 1n the model without having to peruse all
the available detailed material. Since this appendix 1s not Intended as a
guide to the use of the model, none of the Input or output functions of the
model or data storage manipulations within the model are described.
A. THE BASIC EQUATION ,
The basis for the model 1s the continuity equation, which expresses the
conservation of mass of each pollutant 1n a turbulent fluid 1n which chemical
reactions occur. This equation can be written:
a^ +^^ + i^+^^ j/ ifi\ + j/, !!iU-L/* !!i\
3t 3X 3y 32 = 3X \*H 3X ) 3y \*H 3y / 3Z \*V 3Z /
I I L I I I
Time Advection Turbulent Diffusion
Dependence
+ *i •+ si • (1)
L I L. . I
Chemical Sessions
Reaction
where c^ represents the pollutant concentration and 1s a function of space
(x.y.z) and time (t). This equation describing the dynamic behavior of
reactive pollutants 1s fully three-dimensional. Further examination of the
Adapted from Reynolds, Tesche, and Reid (1978).
A-3
-------
equation Indicates that the following physical and chemical processes are
considered 1n the Airshed Model:
> Pollutant advection. The model can treat a fully three-
dimensional wind field, where u, v, and w are the mean wind
velocity components in the x, y, and z directions, respectively;
> Turbulent diffusion. Pollutant transport resulting from the
influence of atmospheric turbulence is treated through the
use of the eddy diffusivity concept. KH and Ky are the
horizontal and vertical diffusivity coefficients, respectively.
> Chemical reaction. The term R^ represents the net rate at
which pollutant 1 is generated by chemical reactions. The
reaction rate is a function of pollutant concentration,
temperature, and the intensity of ultraviolet radiation.
> Emissions. The spatial and temporal distribution of the
source emissions are treated 1n the term S^. In the case
of large point sources, the total effective plume rise is
calculated to enable the appropriate spatial placement of the
emissions aloft.
In addition, removal of pollutants by surface uptake processes is con-
sidered in the boundary conditions of Eq. (1).
To derive Eq. (1), one must make three assumptions. First, pollutant
transport effects due to molecular diffusion are small relative to those
attributable to turbulent diffusion. Second, pollutant transport due to
turbulence can be adequately parameterized through the use of the eddy
diffusivity concept. Third, turbulent concentration fluctuations have a
A-4
-------
negligible Influence on reaction rates. For a more thorough discussion of
the derivation of Eq. (1), the reader 1s referred to reports by Reynolds
et al. (1973a, 1973b).
B. THE NUMERICAL SOLUTION PROCEDURE
Because Eq. (1) 1s nonlinear, 1t 1s not possible to obtain an analytical
solution. Thus, one must utilize appropriate numerical techniques to find an
approximate solution. To facilitate the application of finite difference
methods, one first performs a change of variable to normalize the vertical
dimension by the distance between the bottom and top of the region. This 1s
accomplished by defining a new Independent variable p as follows:
2 - Hb(x,y.t)
p = Htu,y.tji - Hbu,y,tj
where Hb and Ht are the elevations of the bottom and top of the region,
respectively. Upon performing this change of variable and neglecting
small cross-derivative turbulent diffusion terms* Eq. (1) becomes
(Reynolds et al., 1973a, 1973b):
* apW ap/ +
(2)
where
w . u
w u
ax ax / ~ \ ay ay / * at *
A-5
-------
AH = Ht(x,y,t) - Hb(x,y,t)
As indicated in Eq. (2), the Airshed Model requires values of u, v, and
W in order to carry out pollutant advection calculations. Actually, the
user must specify only the values of u and v 1n each grid cell, and the
Airshed Simulation Program computes appropriate values of W from the wind
continuity relationship and information pertaining to the depth of the
modeling region. To calculate W given values of u and v in each grid cell,
one first writes the continuity relationship:
<3'
Upon performing the same change of variable just described, one obtains:
3(UAH) , 3(VAH) t 3W- o
sx ay SP (4)
where
-- /aHb 3AH\ /aHb + 3AH\
(5)
The value of W used in Eq. (2) is then given by:
W = W - P-^ (6)
Equation (4) can be written in finite difference form and solved subject
to the constraint that fl = 0 at the ground. Equation (6) is then used
to complete the specification of the values of W employed 1n the
numerical solution of Eq. (2).
A-6
-------
It should be noted that reliance on this technique for assuring a mass
consistent wind field should be undertaken with caution. If the u-v wind
field within a layer 1s characterized by excessive divergence, as might
occur following use of a simple Interpolation scheme applied to station
measurements, the divergence must be reduced by employing suitable tech-
niques for "smoothing" the wind field prior to the simulation program.
Otherwise, unreal1st1cally large vertical velocities may result.
Special consideration must be given to the application of finite
difference methods to multidimensional problems. The Airshed Model employs
the so-called method of fractional steps. Applying this technique to the
solution of Eq. (2), one obtains the following governing equations for the
four-step numerical integration procedure:
> Step I
&*1> *&«",> •&(«,,•«§•) • (7)
> Step II
^HC^^VAHC^^JVH^?-) . (8)
> Step III
> Step IV
RjAH (id
A-7
-------
If the modeling region 1s divided Into two layers, then there will be
two sets of Eqs. (7) through (10), differing only in the definition of
AH. the numerical Integration of Eq. (2) for one time interval 1s
approximated in the method of fractional steps by the sequential inte-
gration of Eqs. (7) through (10) for one time interval. By carrying
out this four-step procedure for many time Intervals, one simulates the
time history of pollutant concentrations 1n each grid cell.
As indicated by Eqs. (7) and (8), Steps I and II of the solution
procedure treat horizontal transport due to advection and diffusion.
Step I integrates Eq. (7), which 1s transport of pollutants in the
x-direction, and Step II integrates Eq. (8), transport 1n the y-d1rect1on.
Since Steps I and II are basically the same except for the direction of
integration, only Step I is discussed here. Integration of Eq. (7) is
accomplished by solving the advective part of the equation:
^UHc.) +^uAHc.) = 0 ,
using the SHASTA method developed by Boris and Book (1973). This step
is described 1n detail by Reynolds et al . (1976). The diffusive part of
Eq. (7),
is then solved using a standard explicit finite difference technique.
The pollutant concentrations calculated in Step I are then Input to
Step II, which calculates the effect of transport 1n the y-d1rection.
Equations (7) and (8) cannot be solved without specifying suitable
boundary conditions. At any point on the horizontal boundaries, the
A-8
-------
appropriate constraint 1s dependent on whether the wind 1s flowing Into or
out of the modeling region. When the wind 1s flowing out of the region, the
partial derivative of the pollutant concentration with respect to x (Step I)
or y (Step II) 1s set equal to zero. If the wind 1s flowing Into the region,
then the pollutant concentration must be specified along the upwind boundary.
Step III of the numerical procedure entails the Integration of Eq. (9),
which considers pollutant emissions and vertical advectlon and diffusion.
The equation 1s solved through the use of an Implicit finite difference
technique to eliminate stability constraints on the size of the time step
that might otherwise arise 1f the vertical grid spacing becomes relatively
small at some point 1n a simulation. The computation of the vertical wind
speed W 1s described previously by equations (4) through (6). If pollutants
are advected or entrained in through the top of the modeling region, the
user must specify the concentration at each species at the top of the region.
Otherwise, the partial derivative of the pollutant concentration with respect
o p 1s set equal to zero at the upper boundary. If the modeling region has
been segmented Into layers, then the resulting sets of Eq. (9) are coupled at
the Interface between the layers.
Integration of Eq. (10) to Include the contributions of chemical
reactions 1s the final step of the numerical procedure. A Crank-Micol son
difference scheme 1s employed in Step IV, which yields a set of nonlinear
algebraic equations. A solution to these equations is obtained by using a
Newton iterative procedure. For further discussion of the details of the
finite difference equations, the reader is again referred to the reports
by Reynolds et al., (1973a, 1976), and to KWus et al., (1977).
A-9
-------
C. THE ESTIMATION OF TURBULENT DIFFUSIVITIES
Central to the treatment of turbulent diffusion processes in Eq. (2)
is the estimation of the horizontal and vertical diffusivity coefficients,
K^ and Ky. Because of the empirical nature of the diffusivity concept,
these coefficients are difficult to measure and specify precisely. With
regard to horizontal pollutant transport, advection tends to be significantly
more important than turbulent diffusion processes. This dominating influence
is strengthened because local concentration gradients near point and line
sources cannot be represented in the Airshed Model owing to the use of a
grid with relatively coarse horizontal resolution. Although the horizontal
diffusion terms have a small influence on predicted concentrations, they
have been retained in the model. At present, KM is set to a constant value
2
of 50 m /sec.
In contrast to horizontal transport, turbulent diffusion is frequently
the dominating vertical transport process. Thus, considerable attention has
been given to the development of algorithms for estimating the vertical dif-
fusivity, K... The parameters used to calculate Ky are the stability class,
the ground-level wind speed, the surface roughness, the height of the wind
measurements, and the height of the grid cell. The first step in calcu-
lating the diffusivity is to estimate the Monln-Obukhov length. This is
done by relating the Monin-Obukhov length, L, to the surface roughness, ZQ,
and the stability function, S, using the following expression (Liu et al.,
1976):
(DI - b2|S| + b3S2)
A-10
-------
where
a, =
bo =
0.004349,
0.003724,
0.5034,
0.2310,
0.0325.
This formula 1s a result of the best fit of observational data reported
by Colder (1972). The stability function, S, a numerical representation of
i
the Pasqulll stability category, can be calculated as follows:
(14)
where
S1gn(ce)
1
0
-1
ce>0
co <0
The parameters cw and cfi are the wind speed class and exposure class respectively,
and are defined as follows:
u
cw= *
•jr » 0 $ "r 1 8 m/sec ,
4 , u > 8 m/sec ;
. i "*
ce =
daytime insolation
3 » strong
2 , moderate
1 , slight
0 , heavy overcast day or night ,
-1 , •>£ cloud cover
-2 <|- cloud cover
nighttime cloudiness
If S = 0, the Monin-Obukhov length 1s set to 1.0 x 10 meters.
A-11
-------
Next, the friction velocity, u*, is calculated using the following
equation (Liu et al., 1976 ):
where up denotes the wind speed measured at a reference height, zf, and where
(stable or neutral)
(15)
9 =
F an
* an
1 - * (— )
(J * L '
t *
-------
to generate a theoretical concentration field downwind of a point source.
Optimal control theory techniques have been employed to estimate the dif-
fuslvlty profiles that cause the predicted concentration fields of Eq. (2)
to agree most closely with the theoretical fields. The reader 1s referred
to the reports by Lamb (1976), Lamb, Chen, and Se1nf1eld (1975), and
Lamb et al. (1977) for further details.
Mathematically, the algorithms used to calculate Ky can be expressed
as follows:
«
> Stable conditions
ku*z expf -
i
u*
fz\
«*)
/z\ • (18)
1 + 4.7(f
> Neutral conditions
Kv= T~ (ao + V * V2 + V3 + V4)
for 0 < z <.0.45[-F
Ky = 0.01 m2/sec for z > 0.45 (^-) . (20)
> Unstable conditions
K = kl 7 I O + Q r ^ O
Ny "**•! po 1
+ 64(3c4 - 8?2 + 1) 1
A-13
-------
where
c s ZT- -
and
f = Coriolis parameter,
k = von Karman constant,
z. = inversion height,
v = geostrophic wind component.
The coefficients a., and ^ have the following values:
= 7.396 x 10~4 , BQ = 0.152
= 6.082 x 10"2 . BT = 0.080
= 2.532 , B2 = -0.039
= -1.272 x 10 , 83 = 0.032
= 1.517 x 10 , B = 0.020
As noted earlier, the Airshed Model has the capability of including an
upper layer of grid cells (frequently corresponding to the inversion layer)
within the modeling region. To estimate values for Ky in this layer, vertical
temperature gradient inputs are checked to establish the appropriate stability
class of the layer. The corresponding equation cited above and the actual
height of the grid cell is then used to calculate
A-14
-------
D. THE TREATMENT OF ATMOSPHERIC CHEMISTRY
In the earlier discussion of Eq. (2), 1t was Indicated that chemical
effects are Incorporated Into the model through the term R^, the net rate
of production of pollutant 1 by chemical reactions. The net rate of reaction
can be calculated using a chemical kinetic mechanism, which 1s a set of reac-
tions and rate constants that describe the pertinent atmospheric chemical
phenomena. The main chemistry mechanism employed 1n the Airshed Model 1s
known as the Carbon-Bond Mechanism (CBM), as reported by Whltten and Hogo
(1977) and Whltten, Hogo, and KUlus (1979). A more recent version of the
Carbon-Bond Mechanism (CBM II) 1s described 1n Whltten, Klllus, and Hogo
(1980) and Whltten, et al. (1979). The 65 reactions and rate constants in
this mechanism are presented 1n Table A-l.
A unique feature of the Carbon-Bond Mechanism 1s Its treatment of
organic species. Since every reaction of all the organic species found 1n
an urban atmosphere cannot be considered, these pollutants must be grouped
to Hm1t the number of reactions and species to a manageable level while
permitting reasonable accuracy 1n predicting ozone formation. In the
Carbon-Bond Mechanism, each carbon atom of an organic molecule 1s classified
according to Its bond type. For example, the carbon bonds in a butene
molecule (C^Hg) can be represented as follows:
H H
I I
H-C-C-C-C-H
l I I l
H H H H
This molecule has two carbon atoms associated solely with single bonds and
two atoms associated with a double bond. The two carbon atoms associated
with a double bond are also each associated with a single bond. In assigning
A-l 5
-------
TABLE A-1. THE CARBON-BOND MECHANISM (CBM-II)
(WMtten, Kill us, and Hogo, 1980)
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
Reaction
N02 + hv* NO + 0
0 + 02 + M-.03 + M
03 + NO •*> N02 + 02
°3 + N02 * N03 + °2
0 + N02 •* NO + 02
03 + OH -^ H02 + 02
03 + H02 -•• OH + 202
N02 + OH * HN03
09
CO + OH + H02 + C02
NO + NO + 02 •*• 2N02
N03 + NO - 2N02
N03 + N02 + H20 + 2HN03
H02 + NO -»• N02 + OH
H02 +H02 -
PAR + 0 -»• ME02 + OH
PAR + OH + ME02
OLE * 0 •*• ME02 + AC03 + X
OLE + 0 •*• CARB
OLE + OH + RA02
OLE + 03 •»• CARB + CRI6
OLE + 03 * CARB + MCR6
ETH + 0 * ME02 + H02 + CO
ETH + 0 •* CARB
Rate Constant
at 298°K
(ppm win )
Experimental
,§
2.1 x 10'*
23.9
4.8 x 10"2
1.34 x 104
7.7 x 101
5.0
1.4 x 104
2
4.4 x 10^
7.1 x 10"10
2.8 x 104
.***
311 x k (N20g + H20)
1.2x104
1.5 x 104
2 x 101
1.5 x 103
2.7 xlO3
2.7 x 103
4.2 x 104
o
8 x NT*
*
8 x 10 *
6 x 102
6 x 102
Activation
energy
(K)
—
—
1,450
2,450
—
1,000
1,525
—
—
—
—
-10.600
—
--
—
—
—
—
—
— —
—
—
—
A-16
-------
TABLE A-1. (Continued)
Reaction
Rate Constant Activation
at 298°K , energy
(ppm nrln )
**
/
2
24. ETH 4- OH •*• RB02 1.2 x 104
25. ETH + 03 * CARB + CRIG 2.4 x 10"3
26. AC03 4- NO •»• N02 + ME02 4- C02 3.8 x 103
27. RB02 4- NO -» N02 + 2 CARB 4- H02 1.2 x 104
28. RA02 + NO * N02 + 2 CARB + H02 1.2 x 104
29. ME02 + NO •*• N02 + CARB + MEOg + X (1.2 x 104)(A-1)/A*
30. MEO + NO + N02 4- CARB + H02 (1.2 x
31. ME02 + NO •*• Nitrate 5 x 10'
32. RB02 + 03 -»• 2 CARB + H02 5.0
33. RA02 + 03 * 2 CARB + H02 2 x 102
34. ME02 + 03 -»• CARB + H02 5.0
35. CARB •*• OH •»• o(H02 + CO) +
(1 - o) (AC03 4- X) (2.4 - a) x 104
36. CARB 4- hv -»• CO okf*tt
37. CARB 4- hv -»• (1 4- 0)H02 + a + 1 *tf
(1 - o)(ME02 4-X) 4- CO Z f
38. X 4- PAR -v 1 X 105
39. AC03 4- N02 * PAN 2 x 103
40. PAN * AC03 4- N02 2.8 x 10"2t 12,500
41. AC03 + H02 * 4 x 103
42, ME02 4- H02 -»• 4 x 103
43. CRIG + NO •»• N02 4- CARB 1.2 x 104
44. CRIG 4- N02 - N03 4- CARB 8 x 103
45. CRIG 4- CARB * Ozonlde 2 x 103
46. MCRG 4- NO •»• N02 + CARB 1.2 x 104
A-17
-------
TABLE A-l. (Continued)
Reaction
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
MCRG + N02 -* N03 + CARB
MCRG + CARB -» Ozonide
CRIG H- CO
CRIG -»• Stable Products
CRIG -» 2H02 + C02
MCRG -»• Stable Products
MCRG ->• ME02 + OH + CO
MCRG -»• MEO« + HO, + CO,
Z Z Z
MCRG -»• CARB + 2H02 + CO
ARO + OH -»• ARPI + ARPI + ARPI + HOg
ARO + OH + H02 + GLY + X
ARO + OH -»- OH + GLY + W
W + CARB •>
ARPI + NO -»• NO + CARB + PAR
ARPI + NO -»• N02 + AEROSOL
ARPI + N03 -> CARB + CARB
ARPI + 03 •»• AEROSOL
GLY + OH •*• H02 + ARPI + ARPI + ARPI + CO
GLY -» ME02 + H02 + ARPI + ARPI + ARPI
Rate
at
(ppm
8 x
2 x
6.7
2.4
9 x
1.5
3.4
4.25
8.5
6 x
1.6
1.5
1.0
30
15
3.5
0.6
104
*V*I \J
vlLY
Constant
298° K
-1 mm'1)
3
3
x 102t
xl02t
10lf
x!02t
x 102t
x!02t
x 10U
103
x 103
x 104
x 105
x 104
.11
Activation
energy
(K)
—
--
--
—
—
—
«
_ _
—
--
—
—
—
—
—
—
—
—
-—
The rate constants shown are as used to model eleven experiments at UCR that
used mixes of seven hydrocarbons. For that study the default values, a = 0.5
and A = 1.3, were used.
fUn1ts of mln'1.
§
Un1ts of
A-18
-------
TABLE A-1. (Concluded)
**
A « A 1s the average number of R0«-type radicals generated from a hydrocarbon
between attack by OH* and generation of H0«.
fta 1s the fraction of total aldehydes that represent formaldehyde and ketones.
kr Is the carbonyl photolysis rate constant.
C 1 1
n j. u n^ B 5 x 10 ppm" m1n~ for UCR simulations.
'
2 5
A-19
-------
carbon atoms to bond categories 1n the CBM, precedence 1s given to double
bonds and carbonyl bonds. The kinetic mechanism given 1n Table A-l considers
five possible types of carbon bonds: single-bonded carbon atoms (PAR), very
reactive double bonds (OLE), aromatic rings (ARO), carbonyl bonds (CARB), and
moderately reactive double bonds (ETH). Single-bonded carbon atoms comprise
not only paraffin molecules, but also portions of olefln, aromatic, aldehyde
and other molecules. Double bonds are treated as pairs of carbon atoms. An
activated aromatic ring 1s treated as a unit of six carbon atoms. The carbon
atom in the carbon-oxygen double bond of an aldehyde or ketone 1s Included
1n the carbonyl group. To Illustrate this treatment of organic species,
1 mole of butene 1s considered to consist of 2 moles of single-bonded carbon
atoms and 1 mole of very reactive double bonds. Table A-2 shows how other
organic species are treated 1n the Carbon Bond Mechanism.
CBM II exhibits several Important characteristics which make 1t suitable
for incorporation 1n the Airshed Model. First, 1t has no adjustable parameters.
In the formulation of CBM II, special care was taken to avoid the use of
adjustable parameters that might have a profound effect on the predicted results.
Other proposed mechanisms that do contain such parameters can sometimes be
"tuned" to predict well for one set of conditions but must be "retuned" to fit
other conditions. This limits their overall utility.
Second, CBM II was developed to be a condensation of larger, more
detailed mechanisms for the reactions of propylene, butane, ethylene, toluene,
formaldehyde, and acetaldehyde. Tests Indicate that CBM II generates pre-
dictions that closely agree with those obtained from the more detailed
mechanisms. Moreover, this correspondence was obtained without any parameter
adjustments. Thus, CBM II 1s a condensation of present knowledge of the
A-20
-------
TABLE A-2.
SPECIES CHEMICAL RAHE
RO.
1 METHANE
2 ETHANE
3 ETHYLENE
4 PROPANE
6 PROPYLENE
6 ACETYLENE
7 CYCLOPROPANE
8 PROPADIENE
9 METHYLAGETYLENE
10 'CYCLOPENTARE
11 R-BUTANE
12 BUTENE
13 ISO-BUTANE
14 1.3-BDTADIENE
18 ETHYLACETYLENE
16 N-PENTANE
17 l-PENTENE
18 2-METHYL-2-BUTENE
19 HEXARE
20 HEPTANE
21 OCTANE
22 NONANE
23 'ISOMERS OF HEXARE
24 'ISOMER8 OF HEPTANE
26 R-DECARE
26 'I80MER8 OF OCTANE
27 'CYCLOHEXARE
28 UROECARE
29 olSOMERS OF RORARE
30 'I80MER8 OF DECARE
31 "ISOMERS OF URDECARE
32 'R-DODECARE
33 'ISOMERS OF DODECARE
34 'R-TRIDECARE
38 'I80MER8 OF TRIDECARE
36 'R-TETRADECANE
37 'ISOHERS OF TETRADECARE
38 'R-PERTADECARE
39 "1SONERS OF PERTADECARE
40 'C-7 CYCLOPARAFFIRS
41 'C-B CYCLOPARAFFIRS
42 *C-9 CYCLOPARAFFIRS
43 'TERPERES
44 'METHYLCYCLOHEXARE
48 'MIRERAL SPIRITS
CARBON-BOND PROFILES BY COMPOUND (Whltten, 1979)
OLE PAR ARO GARB
e.ee
e.ee
e.ee
e.ee
i.ee
v • OO
v • Vv
e.ee
e.ee
e.ee
e.ee
i.ee
e.ee
i.ee
e.ee
e.ee
i.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
i.ee
e.ee
e.ee
e.ee
e.ee
e.ee
i.8o
i.ee
e.ee
i.ee
e. ee
i.Bo
3.ee
4.ee
2.ee
4.ee
e.ee
i.ee
s.ee
3.ee
3.ee
e.ee
7.ee
a.ee
9.ee
e.ee
7.ee
ie.ee
a.ee
4.ee
n.ee
9.ee
ie.ee
ii.ee
i2.ee
i2.ee
i3.ee
is.ee
I4.ee
i4.ee
is.ee
i8.ee
s.ee
.00
,oe
6.00
6.00
6.00
6,
7,
OAA
• W
e.ee
e.ee
e.ee
e.ee
e.ee
e.eo
e.eo
e.eo
e.ee
o.oe
e.eo
e.ee
e.ee
e.ee
e.ee
e.oe
e.ee
e.oo
e.eo
o.eo
o.eo
e.ee
fr.ee
e.oe
e.ee
e-.ee
e.oe
e.oe
e.oo
o.eo
e.oe
e.eo
e.oe
e.oe
o.oo
e.ee
e.ee
o.oe
e.oe
e.oe
o.ee
o.oo
o.ee
o.oo
e.ee
e.oe
e.ee
e.ee
e.ee
e.ee
e.ee
i.ee
e.ee
e.ee
e.eo
o.oe
e.ee
2.00
e.ee
e.ee
e.eo
2.00
e.oe
e.ee
e.ee
o.oe
e.ee
e.ee
e.ee
e.oo
e.oo
e.oe
e.oo
e.eo
eika
. w
0.00
e.oe
e.eo
e.oe
e.oo
o.oo
e.ee
e.oo
o.eo
2.00
e.ee
e.oo
ETH
e.eo
o.oe
i.eo
e.oe
e.oe
e.ee
i.oo
i.oe
o.oo
i.ee
e.oo
e.oo
e.ee
e.ee
e.ee
e.oe
e.oo
e.ee
o.oo
o.oo
e.oe
o.oo
e.ee
e.ee
e.oe
o.oe
i.oo
e.oo
e.eo
e.ee
o.oo
e.eo
e.oo
o.oo
e.oe
e.oe
e.oo
o.eo
e.ee
oe
oe
oo
o.eo
.00
.00
URREACTIVE
1.00
2.00
e.eo
1.80
e.ee
i.ee
e.ee
e.ee
i.se
e.ee
e.ee
o.oo
o.eo
e.eo
e.oo
e.eo
e.eo
e.ee
o.oo
e.eo
o.ee
e.eo
e.oo
e.eo
e.oe
e.eo
e.oo
o.oo
o.eo
o.oe
o.oo
o.oe
o.oo
o.oe
o.oo
o.eo
o.oo
o.oo
e.oo
o.oo
e.oo
e.oo
o.oo
e.ee
e.oo
-------
TABLE A-2 continued
NJ
K>
SPECIES
NO. CHEMICAL NAME
46 'CYCLOHEXAHORE
47 *LACTOL SPIRITS
48 'I8OHEBS OF BOTIHC
49 'I80MER8 OF PElfTENE
86 'ISOMERS OF PERTAHE
81 METHYL ALCOHOL
82 ETHYL ALCOHOL
83 H-PROPYL ALCOHOL
84 I80-PROPYL ALCOHOL
88 H-BUTYL ALCOHOL
86 ISO-BUTYL ALCOHOL
87 BUTYL CBLL08OLVE
SB TEAT-BUTYL ALCOHOL
89 METHYL CELLO80LVE
60 CELL080LVE
61 D1ACETOHE ALCOHOL
62 ETHYL ETHER
63 *CLYCOL ETHER
64 'CLYCOL
65 'PROPYLENE GLYCOL
66 ETHYLENE GLYCOL
67 TETRAHYDROFURAN
68 ACETIC ACID
69 METHYL ACETATE
76 ETHYL ACETATE
71 PROPYL ACETATE
72 N-BUTYL ACETATE
73 ETHYL ACRYLATE
74 CELLOSOLVE ACETATE
78 *ISOPROPYL ACETATE
76 "METHYL AMYL ACETATE
77 *ISOBOTYL ACETATE
78 DIMETHYL FORMAMIDE
79 "ISODUTYL ISOBUTYRATE
00 FORMALDEHYDE
81 ACETALDEHYDE
82 *BUTYRALDEHYDE
83 ACETONE
84 METHYL ETHYL KETONE
83 METHYL N-BUTYL KETONE
86 METHYL I80BUTYL KETONE
87 ETIIYLENE OXIDE
88 'PROPYLENE OXIDE
89 ACETONITRILE
90 ACRYLONITRILE
OLE
0.00
V • Vtr
9.99
e.ee
0.00
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
i.*e
e.ee
e.oo
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
0.00
PAR
3.ee
O* vV
2.00
3.00
5.00
i.ee
2.00
3.60
3.66
4.66
4.66
8.66
9.V6
2.66
9* 99
8.66
3.66
1.66
1.66
2.66
1.66
i.ee
2.66
e.ee
3.66
4.66
8.66
2.66
4.00
5.66
a.ee
6.00
e.ee
7.66
e.ee
i.ee
3.66
2.66
3.66
0.00
5.00
0.ee
2.00
.66
.66
g
1
ARO
6.66
6.66
6.66
e.ee
e.ee
6.66
e.ee
6.66
e.ee
e.ee
e.ee
6.66
ff.ee
6.66
6.66
6.66
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
CARB
1.66
6.66
2.66
2.66
6.66
6.66
6.66
6.66
6.66
6.66
6.66
1.66
i!ee
1.66
1.66
1.66
i.ee
1.66
i.ee
1.66
i.ee
e.ee
e.ee
i.ee
i.ee
1.66
1.66
2.00
e.ee
e.ee
e.ee
e.ee
ee
ee
ee
ee
ee
66
66
66
6.66
6.66
6.66
6.66
ETH
1.66
e.ee
e.ee
e.ee
e.oe
e.ee
e.ee
e.ee
6.66
6.66
6.66
6.66
6166
6.66
6.66
6.66
e.ee
e.ee
6.66
e.ee
e.ee
i.ee
e.ee
e.ee
e.ee
e.ee
e. ee
i.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
6.66
1.66
UNREACTIVE
6.66
6.66
6.66
e.ee
6.66
e.ee
6.06
6.66
e.ee
e.ee
6.66
6.66
6.«6
6.66
6.66
6.66
e.ee
6.66
6.66
6.66
6.66
6.66
e.ee
s.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
3.00
e.ee
e.00
o.oo
0.00
0.00
o.oo
0.00
e.oo
2.eo
I.ee
2.00
0.66
-------
TABLE A-2 continued
SPECIES
NO.
CHEMICAL NAME
OLE
PAR
ARO
CARB
ETH
UNREACTIVE
91
92
93
94
95
96
97
98
99
100
101
102
103
104
100
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
1212
123
124
120
126
127
128
130
ETHYLAMIRE 0.00
TRIHETHYL AMIIfE . 0.00
METHYL CHLORIDE 0.00
DICHLOROHETHARE 0.00
CHLOROFORM o.oo
CAJ1UON TETRABROMIDE 0.00
'FREOH 11 0.00
ETHYL CHLORIDE 0.00
1.l-DICnLOROETHANE 0.00
1.1.l-TRICHLOROETHAHE 0.00
ETHYLERE DICHLORIDE 0.00
*FR£OR 12 0.00
PERCHLOROETHYLENE 0.00
METBYLERE BROMIDE 0.00
1.1.2-TRICHLOROETBARE 0.00
'FREOR 113 0.00
"TRIMETHYLFLUOROSILAKE 0.00
'MOROCHLORBERZEHE 0.00
VINYL CHLORIDE 0.00
HAPTHA 0.00
BENZENE 0.00
TOLUENE 0.00
ETHYLBENZEHE 0.00
1.3.5-TRIMETHYLBERZERE 0.00
STYRENE 0.00
A-METHYLSTYRERE 0.00
"ISOMERS OF XYLENE 0.00
'DIMKTHYLETHYLBERZERE 0.00
'1.2.3-TRIMETHYLDENZENE 0.00
'180MER9 OF ETHYLTOLOERE .00
'I80MER9 OF BUTYLBERZERE .00
'I8OMERS OF DIETBYLBERZERE
'I80HER8 OF TRIHETBYLBEHZERC
'ISOMER9 OF PROPYLBERZERE
PHENOLS
'XYLENE BASE ACIDS
CHLOROBERZERE
'1.4-DIOXARE
2-ETHOXYETHYL ACE1ATE
TRICHLOROETHYLENE
1.00
3.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
2.00
0.00
0.00
0.00
0.00
0.00
e.oo
8.00
0.00
1.00
2.00
3.00
0.00
1.00
2.00
4.00
3.00
3.00
4.00
4.00
3.00
3.00
0.00
2.00
0.00
2.00
4.0
2.0
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
.00
.00
.oe
.00
.00
.00
.00
.00
.00
.00
.00
.00
0.00
1.00
0.00
0.0*0
i.W
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
2.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
.00
.00
0.00
0.00
0.00
0.00
0.00
0.00
1.00
0.00
0.00
0.00
0.00
0.00
1.00
1.00
0.00
o.oe
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
r.w
1.00
0.00
.00
.00
.00
.00
.00
2.00
2.00
2.00
2.00
1.00
0.00
1.00
2.00
2.00
3.00
6.00
0.00
0.00
6.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
6.00
0.00
6.00
0.00
l:tt
-------
details of several relatively simple, yet representative, hydrocarbon
systems. As future detailed studies of other organic compounds lead to
a better understanding of the smog formation process, CBM II may modified
to incorporate these findings.
It is well known that reaction rate constants are a function of temp-
erature. Thus, the model treats their effects on the values of the non-
photolysis reaction rate constants. In general, the atmosphere is expected
to be more "reactive" on hot days than on cold days. Smog chamber experi-
ments run at 101°F generate about 100 percent more ozone than similar
experiments conducted at 61°F (Pitts et al., 1977). Given the value of a
rate constant, Kr, at some reference temperature, T , the Arrhenius relation-
ship for the value of k at some other temperature T can be written as:
k = k exp
where Ea is the activation energy of the reaction and R is the gas constant.
a
It should be noted that the Arrhenius relationship predicts a smaller temp-
erature effect on ozone production than that actually observed in smog chamber
experiments. Thus, caution should be exercised when applying the model to
situations where the ambient temperatures significantly exceed 30°C, the
nominal temperature at which most smog chamber experiments are performed.
There are four photolysis reactions 1n the CBM, and the rate constant
for each of these reactions is a function of the solar UV radiation Intensity.
Thus, the Airshed Model requires that the user specify the temporal variations
A-24
-------
of these rate constants. Hourly values of these parameters can be estimated
from actual solar radiation measurement data using a relationship such as that
proposed by Schere and Oemerjlan (1977).
Appreciable quantities of aerosols can cause light scattering effects
that alter the Intensity of UV radiation present 1n the atmosphere. As a result,
the photolysis rate constant near the ground can be approximately 30 percent
smaller than Its value at the top of the modeling region when significant
amounts of aerosol are present. Using predictions of aerosol concentrations In
a column of grid cells, the model estimates the rate constants at the bottom
and top of the modeling region. The values at Intermediate points are esti-
mated by linear Interpolations. This 1s described further 1n KWus, et al.
(1977).
E. THE TREATMENT OF EMISSIONS
The Airshed Model can accommodate emissions both at the ground and aloft.
The emissions from a particular source are Injected Into the appropriate
grid cell. Because of the finite resolution of the model, these emissions
are mixed Instantaneously throughout the cell. The magnitude and spatial and
temporal distribution of emissions must be specified for all pollutants being
simulated In the model. Furthermore, emissions of organic compounds must be
grouped according to the carbon-bond categories discussed previously. This
step can be accomplished through the use of available emissions composition
measurements or the results of other studies carried out 1n similar urban
areas. An example of one such study for Los Angeles 1s reported by Trijonis
and Arledge (1976). A large organic species data base 1s reported by Bucon,
Macko, and Taback (1978), also based on Los Angeles.
A-25
-------
Such data can be readily used to apportion hydrocarbon emissions to
the five carbon-bond categories. The basic approach is to first compute
the number of moles of each organic species emitted by a given source as
follows:
where Q = total hydrocarbons emitted (grams)
X.j = weight percent of species i in emissions Q
M.J = molecular weight of species i (grams)
Q_ = moles of species i emitted (gram-moles)
m1
One then uses the carbon-bond profiles by species given in Table A-2 to
compute the number of moles of each carbon-bond associated with each organic
species. For example, if a source emits 120 moles (Q_ ) of cyclohexane
m1
(entry 27 in Table A-2), this is equivalent to emitting 480 moles of single-
bonded carbon atoms (PAR) and 120 moles of moderately reactive double bonds
(ETH). Once the number of moles of carbon-bonds has been computed for
each species, the results can be summed to compute the total number of
noles of each carbon-bond emitted by the source.
The overall procedure is illustrated in Table A-3. In this example, the
weight percents were obtained from Bucon, Macko, and Taback (1978) and are
representative of architectural surface coating emissions, primarily trade
paints. The computations indicate that for every 100 grams of hydrocarbons
emitted, there are 4.66 gram-moles of single-bonded carbon atoms (PAR), 0.121
g-am-moles of aromatic rings (ARO), 0.25 gram-moles of carbonyl bonds (CARB),
and 0.25 gram-moles of moderately reactive double bonds (ETH). To obtain the
Note that the ring structure of cycl oparaf f 1 ns 1s more photochemical ly
reactive than straight chain paraffins. The opening of the ring structure
is treated as 1f the ring 1s a moderately reactive double bond.
A-26
-------
TABLE A-3. EXAMPLE COMPUTATION OF THE EMISSIONS OF THE FIVE CARBON-BONDS FOR AN INDIVIDUAL SOURCE
SPECIES
I
ro
N-Hexane 86.2
Cyclohexane 84.2
Isomers of Xylene 106.2
Toluene 92.1
Ethyl Benzene 106.2
Acetone 58.1
Methyl Ethyl Ketone 72.1
Methyl N-Butyl Ketone 100.2
Methyl Isobutyl Ketone 100.2
Methyl Alcohol 32.0
Ethyl Alcohol 46.1
Isopropyl Alcohol 60.1
N-Butyl Alcohol 74.1
Isobutyl Alcohol 74.1
Propylene Glycol 76.0
Ethylene Glycol 62.1
N-Butyl Acetate 116.2
Isobutyl Acetate 116.2
Dimethyl Formamide 73.1
Isobutyl Isobutyrate 144.2
2-Ethoxyethyl Acetate 132.0
TOTAL
WEIGHT PERCENTS
Weight
Percent
(*)
20.7
20.7
2.6
5.2
4.3
3.2
5.6
0.7
0.6
3.9
0.6
16.4
1.6
0.6
0.8
0.6
2.5
1.5
0.5
6.1
1.3
Compound
(moles)
0.240
0.246
0.024
0.056
0.041
0.055
0.078
0.007
0.006
0.122
0.013
0.273
0.022
0.008
0.011
0.010
0.022
0.013
0.007
0.042
0.010
x 28
PAR
(moles)
1.440
0.984
0.048
0.056
0.082
0.110
0.234
0.035
0.030
0.122
0.026
0.819
0.088
0.032
0.022
0.010
0.110
0.078
0.294
0.040
4.660
x 14
ARO
(moles)
_
-
0.024
0.056
0.041
-
-
-
-
-
-
-
-
—
—
-
-
-
»
-
0.121
x 78
CARB
(moles)
-
-
-
-
0.055
0.078
0.007
0.006
-
_
-
-
-
o.on
0.010
0.022
-
0.042
0.020
0.251
x 30
ETH
(moles)
.
0.246
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
_
-
0.246
x 28
65.2
9.4
7.5
6.9
-------
corresponding weight percent of each carbon-bond, the number of gram-moles 1s
multiplied by the weight of the carbon-bond unit. These weights are as follows
PAR @ CH2 14 grams
OLE @ C2H4 28 grams
ETH @ C2H4 28 grams
ARO @ CgHg 78 grams
CARB @ COH2 30 grams
In this example, the carbon-bond weight percents are therefore 65.2% PAR,
9.4% ARO, 7.5% CARB, and 6.9% ETH. Once such a profile of carbon -bond weight
percents 1s developed, 1t can be readily applied to an emission rate Q to
obtain the number of moles for each carbon-bond as follows:
PAR = (wt^0PAR) Q/14
OLE = () Q/28
ETH = f±fapL) Q/28
ARO = ("Sflft^0) Q/78
CARB = (WtiQQCARB) Q/30.
Emissions released at or near ground level are injected Into the lowest
level of grid cells. Emissions emanating from elevated point sources are
Injected into the appropriate grid cell corresponding to the total effective
plume rise, Ah :
Ahp = Ahs + Ah,,, ,
where Ahe 1s the stack height and Ahm 1s the ultimate plume rise relative
s *
to the top of the stack. To estimate Ah^, the following algorithms recom-
mended by BHggs (1971) are employed:
A-28
-------
> When s > 0 and u >. 1 m/sec,
** - 8.7 - (23)
> When s > 0 and u £ 1 m/sec and when Ah^ is less than
Ah predicted by Eq. (23),
Ah
00
- 8.6 Q1/4 s-3/8 . <24)
> When s < 0 and Q >. 20 MWf
Ah - 15.3 Q1/3 Ah*/3 iT1 . (25)
oo 5
> When s <. 0 and Q < 20 MW,
Ah * 20.8 Q1/3 X2/3 U'1 . (26)
oo
In Eqs. (23) through (26),
s = I-?-1 I-57
X =
and
Q = rate of heat emissions (MW),
u = wind speed at height Ah (m/sec),
g = gravitational acceleration (m/sec ),
T = mean ambient temperature (°K),
e = potential temperature (°K).
A-29
-------
The problem of plume penetration Into or through stable layers aloft
(temperature Inversions) 1s also Important. Normally this 1s handled by
comparing the potential temperature of the plume at some elevation with
that of the ambient air at the same elevation (Briggs, 1971). The plume
should continue to rise as long as Its potential temperature 1s higher.
Therefore, to determine the fate of the plume when it encounters a tempera-
ture inversion, provisions have been made in the Airshed Model to compute
the difference between the plume and ambient potential temperatures, A6,
using relationships suggested by Briggs (1971). If Ae is predicted to be greater
than zero at the base of the Inversion, then the plume 1s allowed to rise until
the point where A6 = 0 is reached. If the plume rises higher than the top of
the modeling region, then the pollutants carried in the plume are excluded
from subsequent Airshed Model calculations.
F. THE TREATMENT OF SURFACE UPTAKE
Pollutant removal processes are Incorporated Into the Airshed Model through
tne use of the deposition velocity concept, that 1s, the uptake flux of pollutant
1 at the surface, F^.,, is proportional to the ground-level concentration
C ,. The proportionality constant 1s the effective deposition velocity,
V.j. Mathematically, the flux can be written as follows:
Fd1 • Vd1 Cg1 (27)
The uptake of a pollutant takes place in three sequential stages:
> Transport to just above the surface by advectlon and
turbulent diffusion.
> Transport to the surface Itself by processes that are
Influenced by the shape of the surface.
> Absorption, adsorption, or chemical reaction at the surface.
A-30
-------
Thus, the overall removal rate Is affected by both transport and chemical
processes.
To estimate values for V,.,, one applies the concept of a resistance to
transport, Rt, and to surface removal, Rg^. One then defines V^ as follows
When either resistance 1s large, V^. approaches zero.
To parameterize the transport resistance, the results of studies
carried out by Owen and Thompson (1963) and Chamberlain (1966) are employed
to obtain:
Rt = u^;2^'^;1 , (29)
where u 1s the wind speed at a reference height of 10 meters, and B
1s an empirically determined function having the following value:
B"1 = 2.2 u;1/3 . (30)
The surface resistance RS^, 1s dependent on the pollutant and on
the type of surface. To determine the surface resistance for grid square j,
, one calculates the average surface uptake velocity V. = (
V
sU • VkVsi . (3D
A-31
-------
where
o.k = fraction of grid square j represented by land use kt
B^ = a factor that adjusts the reference surface uptake
velocity to that for land use k,
^s1 = re^erence surface uptake velocity for pollutant 1.
Values of ek for central business district, suburban residential, and rural/
agricultural land use categories have been estimated at 0.2, 0.5, and 1.0,
respectively (Killus et a!., 1977). To satisfy the input requirements
of the Airshed Model, the user must estimate values of
I ajk8k
for each ground-level grid cell 1n the modeling region. Further information
pertaining to the parameterization of pollutant removal processes, is found
in Killus et al., (1977).
A-32
-------
APPENDIX REFERENCES
Boris, J. P., and D. L. Book (1973), "Flux Corrected Transport. I. SHASTA,
an Algorithm that Works," J. Comp. Phys., Vol. II, pp. 38-69.
Brlggs, G. A. (1971), "Plume Rise: A Recent Critical Review," Nuclear
Safety. Vol. 12, pp. 15-24.
Bucon, H. W., J. F. Macko, and H. J. Taback (1978), "Volatile Organic
Compound (VOC) Species Data Manual," EPA-450/3-78-119, KVB Engineering, Inc.
Tustln, California.
Buslnger, J. A., and S. P. S. Arya (1974), "Height of the Mixed Layer 1n the
Stably Stratified Planetary Boundary Layer," Advances In Geophysics.
Vol. 18A, F. N. Frenklel and R. E. Munn, eds. (Academic Press, New
York, New York).
Chamberlain, A. C. (1966), "Transport of Gases to and from Grass and
Grass-Like Surfaces," Proc. Roy. Soc., Vol. A290, pp. 236-265.
Deardorff, J. W. (1972), "Numerical Investigation of Neutral and Unstable
Planetary Boundary Layers," J. Atmos. Sci.. Vol. 32, pp. 1794-1807.
Golder, D. (1972), "Relations Among Stability Parameters 1n the Surface
Layer," Bound. Layer Meteorol.. Vol. 3, pp. 47-58.
KUlus, J. P., et al., (1977), "Continued Research in Mesoscale A1r Pollution
Simulation Model1ng--Vol. V: Refinements 1n Numerical Analysis, Transport,
Chemistry, and Pollutant Removal," draft report for Environmental Sciences
Research Laboratory, Office of Research and Development, U. S. Environmental
Protection Agency, Contract No. 68-02-2216, Systems Applications, Incor-
porated, San Rafael, California.
Lamb, R. G. (1976), "Continued Research in Mesoscale A1r Pollution Simu-
lation Modeling—Volume III: Modeling of Mlcroscale Phenomena,"
EPA-600/4-76-016c, Systems Applications, Incorporated, San Rafael, California,
Lamb, R. G., W. H. Chen, and J. H. Seinfeld (1975), "Numer1co-Emp1r1cal Analysis
of Atmospheric Diffusion Theories," J. Atmos. Sci.. Vol. 32, pp. 1794-1807.
Lamb, R. G., et al., (1977), "Continued Research 1n Mesoscale Air Pollution
Simulation Modeling—Vol. VI: Further Studies in the Modeling of Micro-
scale Phenomena," draft report for Environmental Sciences Research Labora-
tory, Office of Research and Development, U. S. Environmental Protection
Agency, Contract No. 68-02-2216, Systems Applications, Incorporated, San
Rafael, California.
A-33
-------
L1u, M. K., et al., (1976), "The Chemistry, Dispersion, and Transport of Air
Pollutants Emitted from Fossil Fuel Power Plants 1n California: Data
Analysis and Emission Impact Model," Systems Applications, Incorporated,
San Rafael, California.
Owen, P. R., and W. R. Thompson (1963), "Heat Transfer Across Rough Surfaces,"
J. Fluid Mech.. Vol. 15, pp. 321-334.
Pitts, J. N., et al., (1977), "Mechanisms of Photochemical Reactions 1n
Urban A1r—Volume III: Chamber Studies," EPA-600/3-77-014b, Environ-
mental Protection Agency, Research Triangle Park, North Carolina.
Reynolds, S. D., T. W. Tesche, and L. E. Reid (1978), "An Introduction to
the SAI Airshed Model and Its Usage," draft report to Office of Air
Quality Planning and Standards, U. S. Environmental Protection Agency,
EF78-53R3, Systems Applications, Inc., San Rafael, California.
Reynolds, S. D., et al., (1976), "Continued Research in Mesoscale A1r
Pollution Simulation Modeling: Volume II: Refinements 1n the Treat-
ments of Chemistry, Meteorology, and Numerical Integration Procedures,"
EPA-600/4-76-016b, Systems Applications, Incorporated, San Rafael,
California.
(1973a), "Further Development and Validation of a Simulation
Model for Estimating Ground Level Concentrations of Photochemical
Pollutants," Systems Applications, Incorporated, San Rafael, California.
(1973b). "Mathematical Modeling of Photochemical A1r Pollution—I.
Formulation of the Model," Atmos. Environ.. Vol. 7, pp. 1033-1061.
Schere, K. L., and K. L. Demerjian (1977), "A Photochemical Box Model
for Urban Air Quality Simulations," Proc. of the Fourth Joint Conference
on Sensing of Environmental Pollutants. American Chemical Society.
6-11 November 1977, New Orleans, Louislana.
THjonls, J. C. and K. W. Arledge (1976), "Utility of Reactivity Criteria
in Organic Emission Control Strategies: Application to the Los Angeles
Atmosphere," EPA-600/3-76-091, TRW Environmental Services, Redondo
Beach, California.
Whltten, G. Z., J. P. Killus, and H. Hogo (1980), "Modeling of Simulated
Photochemical Smog with Kinetic Mechanisms, Volume I: Final Report,"
EPA-600/3-80-028a, Systems Applications, Inc., San Rafael, California.
Whltten, G. Z., H. Hogo, and J. P. Kill us (1979), "The Carbon-Bond
Mechanism—A condensed Kinetic Mechanism for Photochemical Smog,"
Submitted for publication to Environ. Scl. Techno!., Systems
Applications, Inc., San Rafael, California.
Whltten, G. Z., et al., (1979), "Modeling of Simulated Photochemical
Smog with Kinetic Mechanisms, Vol. I: Interim Report," EPA-600/3-79-001a,
Systems Applications, Inc., San Rafael, California.
Whltten, G. Z. (1979), Unpublished. Systems Applications, Incorporated,
San Rafael, California.
A-34
-------
Uhltten, G. Z., and H. Hogo (1978), "User's Manual for a Kinetics
Model and Ozone Isopleth Plotting Package," Systems Applications,
Incorporated, San Rafael, California.
Whltten, 6. Z., and H. Hogo (1977), "Mathematical Modeling of Simulated
Photochemical Smog," EPA-600/3-77-001. Systems Applications,
Incorporated) San Rafael, California.
A-35
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA 450/4-80-020
3. RECIPIENT'S ACCESSION-NO.
4. TITLE AND SUBTITLE
Guideline for Applying the Airshed Model to Urban Areas
''RtfcfobeArTE1980
6. PERFORMING ORGANIZATION CODE
'. AUTHOR(S)
8. PERFORMING ORGANIZATION REPORT NO.
^PERFORMING ORGANIZATION NAME AND ADDRESS
ronitoHng and Data Analysis Division
Office of A1r Quality Planning and Standards
Research Triangle Park, North Carolina 27711
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
1
------- |